Article(id=1329456182479139746, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329455965163860100, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202220797, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1662393600000, receivedDateStr=2022-09-06, revisedDate=1684080000000, revisedDateStr=2023-05-15, acceptedDate=null, acceptedDateStr=null, onlineDate=1794896654003, onlineDateStr=2026-11-17, pubDate=1709308800000, pubDateStr=2024-03-02, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794896654003, onlineIssueDateStr=2026-11-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794896654003, creator=13701087609, updateTime=1794896654003, updator=13701087609, issue=Issue{id=1329455965163860100, tenantId=1146029695717560320, journalId=1324759914146529478, year='2024', volume='61', issue='3', pageStart='539', pageEnd='807', issueExtLink='null', onlineDate='null', pubDate='1709308800000', pubDateStr='2024-03-02', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794896602192, creator='13701087609', updateTime=1794896602192, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext=null, issueFiles=null, downloadFileDto=null}, startPage=728, endPage=747, ext={EN=ArticleExt(id=1329456182747575203, articleId=1329456182479139746, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Survey of Automatic Removal of Moiré Pattern, columnId=1327013252946572224, journalTitle=Journal of Computer Research and Development, columnName=Graphics and Image Processing, runingTitle=null, highlight=null, articleAbstract=

Nowadays, digital cameras and smart phones have played an increasingly important role in daily lives, and have become the main tools for people to perceive the world, record information and communicate with each other. When these devices are used to shoot electronic screens, the irregular Moiré patterns in the image are produced due to the overlap of the display devices and digital grids of the camera sensor, which seriously affects the visual quality of the captured images. Therefore, the removal of Moiré pattern is of great significance for the post-processing of captured images. In this paper, recent research about Moiré removal is reviewed in detail, and the existing methods are classified into two categories according to different application scenarios and technical implementations: prior knowledge based Moiré removal and deep learning based methods respectively. According to different training data acquisition and alignment methods, deep learning based Moiré removal techniques can be divided into convolutional neural network (CNN) based methods and generative adversarial network (GAN) based methods. Then, based on the same public dataset, we implement the mainstream deep learning based Moiré removal algorithms, compare and analyze their performance, and summarize the advantages and disadvantages of various methods. Finally, possible directions for future research are discussed.

, authors=Wenfa Qi, Yuxin Liu, Zongming Guo*, authorsList=Wenfa Qi, Yuxin Liu, Zongming Guo, authorCompany=null, correspAuthors=Zongming Guo, authorNote=

born in 1978. PhD, assistant professor. Senior member of CCF. His main research interests include information hiding, image processing, machine learning

born in 1985. Master, engineer. Member of CCF. His main research interests include information hiding, image processing, and machine learning

born in 1966. PhD, professor. Distinguished member of CCF. His main research interests include network video technology, image processing, and information security

, correspAuthorsNote=null, copyrightStatement=Copyright © 2024 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1329456184777618360, articleId=1329456182479139746, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=摩尔纹图案自动去除技术综述, columnId=1327013253131121602, journalTitle=计算机研究与发展, columnName=图形图像处理, runingTitle=null, highlight=null, articleAbstract=

如今,数码相机和智能手机在人们的生活中扮演着越来越重要的角色,已经成为人们感知世界、记录信息和沟通交流的主要工具. 当使用这些设备拍摄电子屏幕时,显示设备和摄像头传感器网格之间往往会发生混叠,通常导致图片中存在不规则分布的摩尔纹干扰图案,从而严重影响了拍摄图像的视觉质量效果. 因此,摩尔纹图案去除方法研究对于拍摄图像的后期处理具有重要意义. 为此,详细梳理了摩尔纹去除研究的发展脉络,并根据不同的适用场景和技术实现将现有方法分为2类:基于先验知识的摩尔纹去除方法和基于深度学习的摩尔纹去除方法. 鉴于深度学习网络中训练数据集的收集和对齐方式不同,该类方法又分为基于卷积神经网络(CNN)的摩尔纹去除方法和基于生成式对抗网络(GAN)的摩尔纹去除方法. 在此基础上,选择相同的公开数据集,对主流的深度学习方法进行算法实现和性能对比分析,并分别总结了各类方法的优缺点. 最后,对未来的研究方向进行展望.

, authors=亓文法, 刘宇鑫, 郭宗明*, authorsList=亓文法, 刘宇鑫, 郭宗明, authorCompany=null, correspAuthors=郭宗明, authorNote=

1978年生. 博士,助理研究员. CCF高级会员. 主要研究方向为信息隐藏、图像处理和机器学习

1985年生. 硕士,工程师. CCF会员. 主要研究方向为信息隐藏、图像处理、机器学习

1966年生. 博士,研究员. CCF杰出会员. 主要研究方向为流媒体技术、图像处理、信息安全

, correspAuthorsNote=
郭宗明()
, copyrightStatement=版权所有 © 2024 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=Uy3sede6XU6Z5NjX6ei4yA==, magXml=Cd0L1URnj+ZL+BBNAQ6/eA==, pdfUrl=null, pdf=Mb4LY64lByfvSAxWNFqPCQ==, pdfFileSize=6787956, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=PqAaOsZ0AMCKVu4Y1/8smA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=CdUMJB2wZZNhSf3Rz0FlLQ==, mapNumber=null, fund=null)}, authors=[Author(id=1329456186623112125, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=qiwenfa@pku.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329456186761524159, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186623112125, language=EN, stringName=Wenfa Qi, firstName=Wenfa, middleName=null, lastName=Qi, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080, bio={"content":"

born in 1978. PhD, assistant professor. Senior member of CCF. His main research interests include information hiding, image processing, machine learning

"}, bioImg=null, bioContent=

born in 1978. PhD, assistant professor. Senior member of CCF. His main research interests include information hiding, image processing, machine learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329456186824438720, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186623112125, language=CN, stringName=亓文法, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1北京大学王选计算机研究所 北京 100080, bio={"img":"ZZXDGMLZQC3DA2r5ZXEr5Q==","content":"

1978年生. 博士,助理研究员. CCF高级会员. 主要研究方向为信息隐藏、图像处理和机器学习

"}, bioImg=ZZXDGMLZQC3DA2r5ZXEr5Q==, bioContent=

1978年生. 博士,助理研究员. CCF高级会员. 主要研究方向为信息隐藏、图像处理和机器学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])]), Author(id=1329456186895741890, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329456186971239364, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186895741890, language=EN, stringName=Yuxin Liu, firstName=Yuxin, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080, bio={"content":"

born in 1985. Master, engineer. Member of CCF. His main research interests include information hiding, image processing, and machine learning

"}, bioImg=null, bioContent=

born in 1985. Master, engineer. Member of CCF. His main research interests include information hiding, image processing, and machine learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329456187055125445, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186895741890, language=CN, stringName=刘宇鑫, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1北京大学王选计算机研究所 北京 100080, bio={"img":"h1q2bVeccs+0iOax62itsQ==","content":"

1985年生. 硕士,工程师. CCF会员. 主要研究方向为信息隐藏、图像处理、机器学习

"}, bioImg=h1q2bVeccs+0iOax62itsQ==, bioContent=

1985年生. 硕士,工程师. CCF会员. 主要研究方向为信息隐藏、图像处理、机器学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])]), Author(id=1329456187113845703, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=guozongming@pku.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329456187185148873, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456187113845703, language=EN, stringName=Zongming Guo, firstName=Zongming, middleName=null, lastName=Guo, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080, bio={"content":"

born in 1966. PhD, professor. Distinguished member of CCF. His main research interests include network video technology, image processing, and information security

"}, bioImg=null, bioContent=

born in 1966. PhD, professor. Distinguished member of CCF. His main research interests include network video technology, image processing, and information security

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329456187252257738, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456187113845703, language=CN, stringName=郭宗明, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=1北京大学王选计算机研究所 北京 100080, bio={"img":"lx625xjoDlabOuL32KTNJQ==","content":"

1966年生. 博士,研究员. CCF杰出会员. 主要研究方向为流媒体技术、图像处理、信息安全

"}, bioImg=lx625xjoDlabOuL32KTNJQ==, bioContent=

1966年生. 博士,研究员. CCF杰出会员. 主要研究方向为流媒体技术、图像处理、信息安全

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])])], keywords=[Keyword(id=1329456187352921035, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, orderNo=1, keyword=Moiré pattern), Keyword(id=1329456187407446988, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, orderNo=2, keyword=image restoration), Keyword(id=1329456187466167245, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, orderNo=3, keyword=multiscale), Keyword(id=1329456187545859022, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, orderNo=4, keyword=deep learning), Keyword(id=1329456187604579279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, orderNo=5, keyword=convolutional neural network), Keyword(id=1329456187688465360, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, orderNo=6, keyword=generative adversarial networks), Keyword(id=1329456187751379921, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, orderNo=1, keyword=摩尔纹图案), Keyword(id=1329456187818488786, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, orderNo=2, keyword=图像恢复), Keyword(id=1329456187885597651, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, orderNo=3, keyword=多尺度), Keyword(id=1329456187948512212, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, orderNo=4, keyword=深度学习), Keyword(id=1329456188019815381, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, orderNo=5, keyword=卷积神经网络), Keyword(id=1329456188078535638, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, orderNo=6, keyword=生成式对抗网络)], refs=[Reference(id=1329456191631111149, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Shuai, Li Chenghua, Nan Nan, et al. MMDM: Multi-frame and multi-scale for image deMoiréing[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway, NJ: IEEE, 2020: 1751−1759, articleTitle=null, refAbstract=null), Reference(id=1329456191719191534, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Niu Wenjia, Zhang Kaihao, Luo Wenhan, et al. Blind motion deblurring super-resolution: When dynamic spatio-temporal learning meets static image understanding[J]. IEEE Transactions on Image Processing, 2021, 30(8): 7101−7111, articleTitle=null, refAbstract=null), Reference(id=1329456191786300399, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C] //Proc of the 18th 2015 Medical Image Computing and Computer-Assisted Intervention. Berlin: Springer, 2015: 234–241, articleTitle=null, refAbstract=null), Reference(id=1329456191840826352, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Fuping, Hu Yang, Liu Weihua, et al. Face impainting algorithm combining face sketch and gate convolution[C]//Proc of 4th Int Conf on Natural Language Processing(ICNLP). Piscataway, NJ: IEEE, 2022: 81−86, articleTitle=null, refAbstract=null), Reference(id=1329456191899546609, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=蒋斌,刘虹雨,杨超,等. 一种基于局部属性生成对抗网络的人脸修复算法[J]. 计算机研究与发展,2019,56(11):2485−2493, articleTitle=null, refAbstract=null), Reference(id=1329456191983432690, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Jiang Bin, Liu Hongyu, Yang Chao, et al. A face inpainting algorithm with local attribute generative adversarial networks[J]. Journal of Computer Research and Development, 2019, 56(11): 2485−2493(in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456192054735859, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Tianyu, Guo Yanqing, Zheng Xin, et al. Moiré pattern removal with multi-scale feature enhancing network[C] //Proc of IEEE Int Conf on Multimedia and Expo Workshops (ICMEW). Piscataway, NJ: IEEE, 2019: 240−245, articleTitle=null, refAbstract=null), Reference(id=1329456192121844724, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Elad M, Aharon M. Image denoising via sparse and redundant representations over learned dictionaries[J]. IEEE Transactions on Image Processing, 2006, 15(12): 3736−3745, articleTitle=null, refAbstract=null), Reference(id=1329456192184759285, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Xiaogang, Kang Singbing, Yang Jie, et al. Fast patch-based denoising using approximated patch geodesic paths[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2013: 1211–1218, articleTitle=null, refAbstract=null), Reference(id=1329456192247673846, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Xu, Qiong Yan, Yang Xia, et al. Structure extraction from texture via relative total variation[J]. ACM Transactions on Graphics, 2012, 31(6): 139−148, articleTitle=null, refAbstract=null), Reference(id=1329456192339948535, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Shou Yuwen, Lin Chinteng. Image descreening by GA-CNN-based texture classification[J]. IEEE Transactions on Circuits and Systems I: Regular Papers, 2004, 51(11): 2287−2299, articleTitle=null, refAbstract=null), Reference(id=1329456192407057400, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Nah S, Kim T H, Lee K M, Deep multi-scale convolutional neural network for dynamic scene deblurring[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017: 3883–3891, articleTitle=null, refAbstract=null), Reference(id=1329456192461583353, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Nishioka K, Hasegawa N, Ono K, et al. Endoscope system provided with low-pass filter for Moiré removal: US, US08/917429[P]. 2000-02-15, articleTitle=null, refAbstract=null), Reference(id=1329456192549663738, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Schoberl M, Schnurrer W, Oberdorster A, et al. Dimensioning of optical birefringent anti-alias filters for digital cameras[C] //Proc of IEEE Int Conf on Image Process (ICIP). Piscataway, NJ: IEEE, 2010: 4305–4308, articleTitle=null, refAbstract=null), Reference(id=1329456192625161211, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Menon D, Calvagno G. Color image demosaicking: An overview[J]. Signal Processing: Image Communication, 2011, 26(8/9): 518−533, articleTitle=null, refAbstract=null), Reference(id=1329456192683881468, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Pekkucuksen I, Altunbasak Y. Multiscale gradients-based color filter array interpolation[J]. IEEE Transactions on Image Processing, 2013, 22(1): 157−165, articleTitle=null, refAbstract=null), Reference(id=1329456192763573245, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=邱菊. 基于插值算法的去彩色摩尔纹效应的研究[D]. 南京:南京理工大学,2009, articleTitle=null, refAbstract=null), Reference(id=1329456192830682110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Qiu Ju. Research on color Moiré removal based on interpolation algorithm [D]. Nanjing: Nanjing University of Science and Technology, 2009 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456192927151103, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=邱香香. 基于CCD静态图像的摩尔纹去除算法研究[D]. 南京:南京理工大学,2012, articleTitle=null, refAbstract=null), Reference(id=1329456192985871360, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Qiu Xiangxiang. Research on Moiré removal algorithm based on CCD static image[D]. Nanjing: Nanjing University of Science and Technology, 2012 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456193061367808, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Hibbard R H. Apparatus and method for adaptively interpolating a full color image utilizing luminance gradients: US, US08/085520[P]. 1995-01-17, articleTitle=null, refAbstract=null), Reference(id=1329456193191391233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Hamilton J F. Adaptive color plane interpolation in single sensor color electronic camera: US, US08/407423[P]. 1997-05-13, articleTitle=null, refAbstract=null), Reference(id=1329456193279471618, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Kimmel R. Demosaicing: Image reconstruction from color CCD samples[J]. IEEE Transactions on Image Processing, 1999, 9(8): 1221−1228, articleTitle=null, refAbstract=null), Reference(id=1329456193342386179, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=刘芳蕾. 基于图像分解的纹理图像摩尔纹消除方法[D]. 天津:天津大学,2017, articleTitle=null, refAbstract=null), Reference(id=1329456193405300740, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Fanglei. Moiré pattern removal from Textured images via image decomposition[D]. Tianjin : Tianjin University, 2017 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456193464020997, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=张雪. 基于图像分解的屏摄图像摩尔纹消除方法[D]. 天津:天津大学,2018, articleTitle=null, refAbstract=null), Reference(id=1329456193556295686, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Xue. Moiré pattern removal from scree-shot images via image decomposition[D]. Tianjin : Tianjin University, 2018 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456193610821639, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Fanglei, Yang Jingyu, Yue Huanjing. Moiré pattern removal from texture images via low-rank and sparse matrix decomposition[C/OL] //Proc of IEEE Int Conf on Visual Communications and Image Processing (VCIP). Piscataway, NJ: IEEE, 2015[2016-08-25].https:// ieeexplore. ieee. org/stamp/stamp. jsp?tp=&arnumber=7457907, articleTitle=null, refAbstract=null), Reference(id=1329456195234017288, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Jingyu, Zhang Xue, Cai Changrui, et al. DeMoiréing for screen-shot images with multi-channel layer decomposition[C/OL] //Proc of IEEE Visual Communications and Image Processing (VCIP). Piscataway, NJ: IEEE, 2017[2018-03-01].https://ieeexplore.ieee.org/stamp/stamp.jsp?tp =& arnumber=8305057, articleTitle=null, refAbstract=null), Reference(id=1329456195322097673, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Fang Faming, Wang Tingting, Wu Shuyan, et al. Removing Moiré patterns from single images[J]. Information Sciences, 2020, 514(8): 56−71, articleTitle=null, refAbstract=null), Reference(id=1329456195393400842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Sasada R, Yamada M, Hara S, et al. Stationary grid pattern removal using 2D technique for Moiré-free radiographic image display[C] //Proc of Conf on Medical Imaging: Visualization, Image-Guided Procedures, and Display. Bellingham, WA: SPIE, 2003: 688–697, articleTitle=null, refAbstract=null), Reference(id=1329456195456315403, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Sidorov D N, Kokaram A C, Suppression of Moiré patterns via spectral analysis[C] //Proc of Visual Communications and Image Processing. Bellingham, WA: SPIE, 2002: 895–906, articleTitle=null, refAbstract=null), Reference(id=1329456195527618572, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Wei Zhouping, Wang Jian, Nichol H, et al. A median-Gaussian filtering framework for Moiré pattern noise removal from X-ray microscopy image[J]. Micron, 2012, 43(2/3): 170−176, articleTitle=null, refAbstract=null), Reference(id=1329456195594727437, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Hazavei S M, Shahdoosti H R. A new method for removing the Moiré pattern from images[J]. arXiv preprint, arXiv: 1701. 09037, 2017, articleTitle=null, refAbstract=null), Reference(id=1329456195649253390, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Garcia D C, Queiroz R L. Face-spoofing 2D-detection based on Moiré-pattern analysis[J]. IEEE Transactions on Information Forensics and Security, 2015, 10(4): 778−786, articleTitle=null, refAbstract=null), Reference(id=1329456195712167951, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Sidorov D N, Kokaram A C. Removing Moiré from degraded video archives[C/OL] //Proc of the 11th European Signal Processing Conf. Piscataway, NJ: IEEE , 2002[2015-03-30].https://ieeexplore.ieee.org/ stamp/stamp.jsp?tp=&arnumber=7071920, articleTitle=null, refAbstract=null), Reference(id=1329456195766693904, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=吕旭,胡柏青,李开龙,等. 非线性目标跟踪滤波算法的比较研究[J]. 火力与指挥控制,2021,46(4):24−30, articleTitle=null, refAbstract=null), Reference(id=1329456195821219857, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Lü Xu, Hu Baiqing, Li Kailong, et al. Comparison of nonlinear target tracking filtering algorithm[J]. Fire Control & Command Control, 2021, 46(4): 24−30(in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456195884134418, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Bin, Li Shutao, Sun Jun. Scanned image descreening with image redundancy and adaptive filtering[J]. IEEE Transactions on Image Processing, 2014, 23(8): 3698−3710, articleTitle=null, refAbstract=null), Reference(id=1329456195947048979, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Dong Chao, Loy Chenchang, He Kaiming, et al. Learning a deep convolutional network for image super-resolution[C] //Proc of the 13th European Conf on Computer Vision. Berlin: Springer, 2014: 184–199, articleTitle=null, refAbstract=null), Reference(id=1329456196001574932, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Kim J, Lee J K, Lee K M, Convolutional networks for biomedical using very deep convolutional networks[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2016: 1646–1654, articleTitle=null, refAbstract=null), Reference(id=1329456196068683797, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Gharbi M, Chaurasia G, Paris S, et al, Deep joint demosaicking and denoising[J], ACM Transaction of Graphics, 2016, 35(6): 1−12, articleTitle=null, refAbstract=null), Reference(id=1329456196144181270, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Kai, Zuo Wangmeng, Chen Yujin, et al. Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3142−3155, articleTitle=null, refAbstract=null), Reference(id=1329456196207095831, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Kai, Zuo Wangmeng, Chen Yujin, et al. Learning deep CNN denoiser prior for image restoration[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017: 2808–2817, articleTitle=null, refAbstract=null), Reference(id=1329456196265816088, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Abraham E. Moiré pattern detection using wavelet decomposition and convolutional neural network[C] //Proc of IEEE Symp Series on Computational Intelligence (SSCI). Piscataway, NJ: IEEE, 2018: 1275−1279, articleTitle=null, refAbstract=null), Reference(id=1329456196328730649, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Yujing, Yu Yizhou, Wang Wenping. Moiré photo restoration using multiresolution convolutional neural networks[J]. IEEE Transactions on Image Processing, 2018, 27(8): 4160−4172, articleTitle=null, refAbstract=null), Reference(id=1329456196387450906, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Bolin, Shu Xiao, Wu Xiaolin. DeMoiréing of camera-captured screen images using deep convolutional neural network[J]. arXiv preprint, arXiv: 1804. 03809, 2018, articleTitle=null, refAbstract=null), Reference(id=1329456196462948379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Tianyu, Guo Yanqing, Zheng Xin, et al. Moiré pattern removal with multi-scale feature enhancing network[C] //Proc of IEEE Int Conf on Multimedia and Expo Workshops. Piscataway, NJ: IEEE, 2019: 240−245, articleTitle=null, refAbstract=null), Reference(id=1329456196534251548, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=He Bin, Wang Ce, Shi Boxin, et al. Mop Moiré patterns using mopnet[C] //Proc of IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2019: 2424−2432, articleTitle=null, refAbstract=null), Reference(id=1329456196613943325, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=49, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Shanghui, Lei Yajing, Xiong Shuangyu, et al. High resolution deMoiré network[C] //Proc of IEEE Int Conf on Image Processing (ICIP). Piscataway, NJ: IEEE, 2020: 888-892, articleTitle=null, refAbstract=null), Reference(id=1329456196672663582, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=50, authorNames=null, journalName=null, refType=null, unstructuredReference=Cheng Xi, Fu Zhenyong, Yang Jian. Multi-scale dynamic feature encoding network for image deMoiréing[C] //Proc of IEEE/CVF Int Conf on Computer Vision Workshop (ICCVW). Piscataway, NJ: IEEE, 2019: 3486-3493, articleTitle=null, refAbstract=null), Reference(id=1329456196727189536, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=51, authorNames=null, journalName=null, refType=null, unstructuredReference=Cheng Xi, Fu Zhenyong, Yang Jian. Improved multi-scale dynamic feature encoding network for image deMoiréing[J]. Pattern Recognition, 2021, 116(1): 1−12, articleTitle=null, refAbstract=null), Reference(id=1329456196790104097, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=52, authorNames=null, journalName=null, refType=null, unstructuredReference=Zheng Bolun, Yuan Shanxi, Slabaugh G, et al. Image deMoiréing with learnable bandpass filters[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2020: 3636–3645, articleTitle=null, refAbstract=null), Reference(id=1329456196844630050, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=53, authorNames=null, journalName=null, refType=null, unstructuredReference=Park H, Vien A G, Koh Y J, et al. Unpaired image deMoiréing based on cyclic Moiré learning[C] //Proc of Asia–Pacific Signal Information Processing Association Annual Summit and Conf. Piscataway, NJ: IEEE, 2021: 146–150, articleTitle=null, refAbstract=null), Reference(id=1329456196899156003, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=49, rfOrder=54, authorNames=null, journalName=null, refType=null, unstructuredReference=Yue Huanjin, Cheng Yijia, Liu Fanglong, et al. Unsupervised Moiré pattern removal for recaptured screen images[J]. Neurocomputing, 2021, 456: 352−363, articleTitle=null, refAbstract=null), Reference(id=1329456196957876260, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=50, rfOrder=55, authorNames=null, journalName=null, refType=null, unstructuredReference=Park H, Vien A G, Kim H, et al. Unpaired screen-shot image deMoiréing with cyclic Moiré learning[J]. IEEE Access,2022,10:16254-16268, articleTitle=null, refAbstract=null), Reference(id=1329456197016596517, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=51, rfOrder=56, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Jingyu, Liu Fanglei, Yue Huanjing, et al. Textured image deMoiréing via signal decomposition and guided filtering[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3528−3541, articleTitle=null, refAbstract=null), Reference(id=1329456197100482598, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=52, rfOrder=57, authorNames=null, journalName=null, refType=null, unstructuredReference=Ok J, Youn S, Seo G, et al. Paper check image quality enhancement with Moiré reduction[J]. Multimedia Tools and Applications, 2017, 76(20): 21423−21450, articleTitle=null, refAbstract=null), Reference(id=1329456197167591463, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=53, rfOrder=58, authorNames=null, journalName=null, refType=null, unstructuredReference=Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[J]. arXiv preprint, arXiv: 1505. 04597v1, 2015, articleTitle=null, refAbstract=null), Reference(id=1329456197238894632, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=54, rfOrder=59, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Huang, Zhuang Liu, Laurens V D M, et al. Densely connected convolutional networks[J]. arXiv preprint, arXiv: 1608. 06993v5, 2018, articleTitle=null, refAbstract=null), Reference(id=1329456197314392105, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=55, rfOrder=60, authorNames=null, journalName=null, refType=null, unstructuredReference=Hu Jie, Shen Li, Sun Gang. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011−2023, articleTitle=null, refAbstract=null), Reference(id=1329456197373112362, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=56, rfOrder=61, authorNames=null, journalName=null, refType=null, unstructuredReference=Huang Xun, Belongie S. Arbitrary style transfer in real-time with adaptive instance normalization[C] //Proc of IEEE Int Conf on Computer Vision (ICCV). Piscataway, NJ: IEEE, 2017: 1501–1510, articleTitle=null, refAbstract=null), Reference(id=1329456197436026923, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=57, rfOrder=62, authorNames=null, journalName=null, refType=null, unstructuredReference=Vien A G, Park H, Lee C. Dual-domain deep convolutional neural networks for image deMoiréing[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway, NJ: IEEE, 2020: 1934−1942, articleTitle=null, refAbstract=null), Reference(id=1329456197494747180, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=58, rfOrder=63, authorNames=null, journalName=null, refType=null, unstructuredReference=Vien A G, Park H, Lee C. Moiré artifacts removal in screen-shot images via multiple domain learning[C] //Proc of Asia-Pacific Signal and Information Processing Association Annual Summit and Conf (APSIPA ASC). Piscataway, NJ: IEEE, 2020: 1268−1273, articleTitle=null, refAbstract=null), Reference(id=1329456197561856045, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=59, rfOrder=64, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Hailing, Tian Qiaoyu, Li Liang, et al. Image deMoiréing with a dual domain distilling network[C/OL] //Proc of IEEE Int Conf on Multimedia and Expo (ICME). Piscataway, NJ: IEEE, 2021[2021-06-09]. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9428091, articleTitle=null, refAbstract=null), Reference(id=1329456197628964910, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=60, rfOrder=65, authorNames=null, journalName=null, refType=null, unstructuredReference=Zheng Bolun, Yuan Shanxi, Yan Chengganng, et al. Learning frequency domain priors for image deMoiréing[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(11): 7705−7717, articleTitle=null, refAbstract=null), Reference(id=1329456197691879471, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=61, rfOrder=66, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Lin, Liu Jianzhuang, Yuan Shanxin, et al. Wavelet-based dual-branch network for image deMoiréing[J]. arXiv preprint, arXiv: 2007. 07173v2, 2020, articleTitle=null, refAbstract=null), Reference(id=1329456197746405424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=62, rfOrder=67, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Huang, Zhuang Liu, Laurens M, et al. Densely connected convolutional networks[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017, 2261−2269, articleTitle=null, refAbstract=null), Reference(id=1329456197809319985, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=63, rfOrder=68, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu F, Koltun V. Multi-scale context aggregation by dilated convolutions[J]. arXiv preprint, arXiv: 1511. 07122, 2016, articleTitle=null, refAbstract=null), Reference(id=1329456197863845938, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=64, rfOrder=69, authorNames=null, journalName=null, refType=null, unstructuredReference=Zheng Yuanjie, Yu Jingyi, Kang Singbing, et al. Single-image vignetting correction using radial gradient symmetry[C/OL] //Proc of IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2008[2008-08-05].https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=& arnumber =4587413, articleTitle=null, refAbstract=null), Reference(id=1329456197989675059, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=65, rfOrder=70, authorNames=null, journalName=null, refType=null, unstructuredReference=Dai Tao, Cai Jianrui, Zhang Yongbing, et al. Second-order attention network for single image super-resolution[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2019: 11065–11074, articleTitle=null, refAbstract=null), Reference(id=1329456198086144052, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=66, rfOrder=71, authorNames=null, journalName=null, refType=null, unstructuredReference=Mao Xudong, Li Qing, Xie Haoran, et al. Least squares generative adversarial networks[C] //Proc of IEEE Int Conf on Computer Vision (ICCV). Piscataway, NJ: IEEE, 2017: 2794–2802, articleTitle=null, refAbstract=null), Reference(id=1329456199696756789, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=67, rfOrder=72, authorNames=null, journalName=null, refType=null, unstructuredReference=Yue Huanjing, Mao Yan, Liang Lipu, et al. Recaptured screen image deMoiréing[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2021, 31(1): 49−60, articleTitle=null, refAbstract=null), Reference(id=1329456199801614390, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=68, rfOrder=73, authorNames=null, journalName=null, refType=null, unstructuredReference=Yuan Shanxin, Timofte R, Gregory S, et al. Aim 2019 challenge on image demoreing: Dataset and study[C] //Proc of IEEE Int Conf on Computer Vision (ICCV) Workshops. Piscataway, NJ: IEEE, 2019: 3526–3533, articleTitle=null, refAbstract=null), Reference(id=1329456199868723256, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=69, rfOrder=74, authorNames=null, journalName=null, refType=null, unstructuredReference=Yuan Shanxin, Timofte R, Leonardis A, et al. NTIRE 2020 challenge on image demoireing: Methods and results[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway, NJ: IEEE, 2020: 1882-1893, articleTitle=null, refAbstract=null), Reference(id=1329456200044884025, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=70, rfOrder=75, authorNames=null, journalName=null, refType=null, unstructuredReference=Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91–110, articleTitle=null, refAbstract=null), Reference(id=1329456200111992890, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=71, rfOrder=76, authorNames=null, journalName=null, refType=null, unstructuredReference=Chin T J, Yu J, Suter D. Accelerated hypothesis generation for multi-structure robust fitting[C] //Proc of the 11th European Conf on Computer Vision. Berlin: Springer, 2010: 533–546, articleTitle=null, refAbstract=null), Reference(id=1329456200174907451, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=72, rfOrder=77, authorNames=null, journalName=null, refType=null, unstructuredReference=高天宇. 面向移动终端的摩尔纹去除系统的设计与实现[D]. 大连:大连理工大学,2021, articleTitle=null, refAbstract=null), Reference(id=1329456200233627708, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=72, rfOrder=78, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Tianyu. Design and implementation of Moiré removal system for mobile terminals[D]. Dalian: Dalian University of Technology, 2021 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329456200309125181, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=73, rfOrder=79, authorNames=null, journalName=null, refType=null, unstructuredReference=Kaur R, Sharma D V. Pre-processing for improving the recognition of mobile based Gurmukhi text recognition system[C/OL] //Proc of the 12th Int Conf on Computing Communication and Networking Technologies (ICCCNT). Piscataway, NJ: IEEE, 2021[2021-11-03]. http:// ieeexplore.ieee.org/stamp/ stamp.jsp?tp=&arnumber=9579867, articleTitle=null, refAbstract=null), Reference(id=1329456200376234046, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=74, rfOrder=80, authorNames=null, journalName=null, refType=null, unstructuredReference=Fang Han, Zhang Weiming, Zhou Huang, et al. Screen-shooting resilient watermarking[J]. IEEE Transactions on Information Forensics and Security, 2019, 14(6): 1403−1418, articleTitle=null, refAbstract=null), Reference(id=1329456200426565695, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=75, rfOrder=81, authorNames=null, journalName=null, refType=null, unstructuredReference=Howard A, Sandler M, Chen Bo, et al. Searching for mobilenetv3 [C] //Proc of IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2019: 1314−1324, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])], figs=[ArticleFig(id=1329456188195976151, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 1, caption=Moiré pattern of different scales, frequencies, and colors, figureFileSmall=gbdEuFfRphU58dZfHgZ5zA==, figureFileBig=PqAaOsZ0AMCKVu4Y1/8smA==, tableContent=null), ArticleFig(id=1329456188271473624, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图1, caption=不同尺度、频率和颜色的摩尔纹, figureFileSmall=gbdEuFfRphU58dZfHgZ5zA==, figureFileBig=PqAaOsZ0AMCKVu4Y1/8smA==, tableContent=null), ArticleFig(id=1329456188485383129, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 2, caption=The architecture of DMCNN[40] model, figureFileSmall=uaS94UgXm54k8mOzskM6tA==, figureFileBig=BdLkw6WVkQScEeQy2Ex0nQ==, tableContent=null), ArticleFig(id=1329456188544103386, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图2, caption=DMCNN[40]模型架构, figureFileSmall=uaS94UgXm54k8mOzskM6tA==, figureFileBig=BdLkw6WVkQScEeQy2Ex0nQ==, tableContent=null), ArticleFig(id=1329456188607017947, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 3, caption=The architecture of HRDN[44] model, figureFileSmall=ctEkLAd9ezJJOHQWCCtkpw==, figureFileBig=gubJ3Nxdo1UZEmv5gUBezw==, tableContent=null), ArticleFig(id=1329456188686709724, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图3, caption=HRDN[44]模型架构, figureFileSmall=ctEkLAd9ezJJOHQWCCtkpw==, figureFileBig=gubJ3Nxdo1UZEmv5gUBezw==, tableContent=null), ArticleFig(id=1329456188753818589, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 4, caption=Architecture of original and of improved MTRB models, figureFileSmall=aEM80VYWPXIh5rpL3yC2Vw==, figureFileBig=2kUuZlaYj4ik1onTCtmq1Q==, tableContent=null), ArticleFig(id=1329456188808344542, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图4, caption=原始的及改进的MTRB模型架构, figureFileSmall=aEM80VYWPXIh5rpL3yC2Vw==, figureFileBig=2kUuZlaYj4ik1onTCtmq1Q==, tableContent=null), ArticleFig(id=1329456188862870495, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 5, caption=The architecture of EEUCN[50], figureFileSmall=+TrWCUjYgcWU1Uut4DqNpQ==, figureFileBig=bHvYbiZWryu2w/P0n+zz4Q==, tableContent=null), ArticleFig(id=1329456188913202144, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图5, caption=EEUCN[50]结构, figureFileSmall=+TrWCUjYgcWU1Uut4DqNpQ==, figureFileBig=bHvYbiZWryu2w/P0n+zz4Q==, tableContent=null), ArticleFig(id=1329456188971922401, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 6, caption=Image acquisition[40], figureFileSmall=QX+LbTxvolMtNIC0dMgaEw==, figureFileBig=/uPdDx4p1rPhOb3mbhWi4w==, tableContent=null), ArticleFig(id=1329456189026448354, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图6, caption=图像采集[40], figureFileSmall=QX+LbTxvolMtNIC0dMgaEw==, figureFileBig=/uPdDx4p1rPhOb3mbhWi4w==, tableContent=null), ArticleFig(id=1329456189085168611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Figure 7, caption=Comparison of results for the validation set in TIP2018, figureFileSmall=aded4ccGDsPWDaFeXoXHZA==, figureFileBig=tvYsWkf/vVngGGSmH6kTjQ==, tableContent=null), ArticleFig(id=1329456189148083172, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=图7, caption=在TIP2018验证数据集上的对比结果, figureFileSmall=aded4ccGDsPWDaFeXoXHZA==, figureFileBig=tvYsWkf/vVngGGSmH6kTjQ==, tableContent=null), ArticleFig(id=1329456189206803429, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Table 1, caption=

Comparative Analysis of the Morié Removal Algorithms Based on Prior Knowledge

, figureFileSmall=null, figureFileBig=null, tableContent=
算法类别算法算法介绍缺点及不足
图像滤波光学低通滤波器[12]用2个透镜将光信号折射至2个方向,从而降低信号源的频率以避免混叠.将摩尔纹当成了一种高频噪声,通过去除噪声
的方式达到摩尔纹去除的目的. 但摩尔纹图案
广泛分布在图像的各个频带,传统的基于图像
滤波的方法并不能完全解决摩尔纹去除的问题.
中值高斯滤波算法[28]使用局部中值滤波器半自动检测傅里叶振幅谱中的谱峰,以及使用高斯陷波滤波器消除谱噪声峰值.
基于非线性滤波器
算法[31]
基于图像傅里叶频谱幅值的阈值化设计非线性滤波器,减少摩尔纹图案的非平稳性而可能出现的振铃伪影的影响.
自适应滤波去网纹[33]实现了从扫描半色调图像中恢复出高质量连续调图像.
图像插值基于图像色差算法[16]将向量的概念引入到插值算法中实现了RGB通道的3维插值,恢复后的图像更加平滑、更接近实际值.算法实现过于复杂,运算效率较低,
尤其难以在移动设备中实现.
基于CCD图像插值
方案[17]
利用自适应方法选择关联度高的颜色分量来协助判断通道插值方向,图像插值结果更为合理.
基于色差梯度的插值算法[18]基于图像在水平和竖直方向上的亮度和色度差计算梯度,然后将得到的梯度值来确定针对亮度和色度进行插值的方向,最后根据方向做插值运算.
图像分解
基于低秩稀疏矩阵
分解模型[23]
对纹理分量添加低秩先验约束,对摩尔纹分量添加稀疏先验约束以及在其频域分布内添加位置约束,从而区分摩尔纹分量和纹理分量.摩尔纹分量与纹理分量在频域存在能量混叠,
因此基于图像分解方法容易将纹理信息误判
为摩尔纹成分,纹理恢复图像中存在轻微的
振铃效应.
基于信号分解与导向滤波算法[51]将文献[23]利用在G通道,再利用RGB颜色间的相关性用导向滤波方法恢复R,B通道.
基于GMM的多相
图层分解模型[24]
引入了2个基于 GMM的图像先验,分别对背景图层和摩尔纹图层进行规则化,基于GMM的图层分解模型最终获得无摩尔纹的屏摄图像.
), ArticleFig(id=1329456190825804774, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=表1, caption=

基于先验知识的摩尔纹去除算法对比分析

, figureFileSmall=null, figureFileBig=null, tableContent=
算法类别算法算法介绍缺点及不足
图像滤波光学低通滤波器[12]用2个透镜将光信号折射至2个方向,从而降低信号源的频率以避免混叠.将摩尔纹当成了一种高频噪声,通过去除噪声
的方式达到摩尔纹去除的目的. 但摩尔纹图案
广泛分布在图像的各个频带,传统的基于图像
滤波的方法并不能完全解决摩尔纹去除的问题.
中值高斯滤波算法[28]使用局部中值滤波器半自动检测傅里叶振幅谱中的谱峰,以及使用高斯陷波滤波器消除谱噪声峰值.
基于非线性滤波器
算法[31]
基于图像傅里叶频谱幅值的阈值化设计非线性滤波器,减少摩尔纹图案的非平稳性而可能出现的振铃伪影的影响.
自适应滤波去网纹[33]实现了从扫描半色调图像中恢复出高质量连续调图像.
图像插值基于图像色差算法[16]将向量的概念引入到插值算法中实现了RGB通道的3维插值,恢复后的图像更加平滑、更接近实际值.算法实现过于复杂,运算效率较低,
尤其难以在移动设备中实现.
基于CCD图像插值
方案[17]
利用自适应方法选择关联度高的颜色分量来协助判断通道插值方向,图像插值结果更为合理.
基于色差梯度的插值算法[18]基于图像在水平和竖直方向上的亮度和色度差计算梯度,然后将得到的梯度值来确定针对亮度和色度进行插值的方向,最后根据方向做插值运算.
图像分解
基于低秩稀疏矩阵
分解模型[23]
对纹理分量添加低秩先验约束,对摩尔纹分量添加稀疏先验约束以及在其频域分布内添加位置约束,从而区分摩尔纹分量和纹理分量.摩尔纹分量与纹理分量在频域存在能量混叠,
因此基于图像分解方法容易将纹理信息误判
为摩尔纹成分,纹理恢复图像中存在轻微的
振铃效应.
基于信号分解与导向滤波算法[51]将文献[23]利用在G通道,再利用RGB颜色间的相关性用导向滤波方法恢复R,B通道.
基于GMM的多相
图层分解模型[24]
引入了2个基于 GMM的图像先验,分别对背景图层和摩尔纹图层进行规则化,基于GMM的图层分解模型最终获得无摩尔纹的屏摄图像.
), ArticleFig(id=1329456190905496551, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Table 2, caption=

Summary and Comparison of Different Training Datasets

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集发布年份图像分辨率(像素)数据规模图像对生成过程
TIP2018[40]2018256×256135 000个图像对,其中90%用于
训练、10%用于验证和测试.
1)图像捕获:将参考图像按照一定的时间间隔在屏幕中显示,用手机拍摄视频,然后提取相应的图像帧;
2)图像对齐:依靠黑色图像边界的角点,计算出投影矩阵,实现图像对齐.
MRBI[67]2021128×128352个训练图像对,340个测试图
像对.
1)图像捕获:利用屏幕截屏和屏幕拍照的方式得到干净图像和摩尔纹图像;
2)图像对齐:基于尺度不变特征找到2个图像中可以匹配的点对,进而得到变换矩阵.
LCD Moiré[68]20191024×102410 200个图像对,其中10 000个用
于训练、100个用于验证及100个用于测试的图像对.
针对干净图像利用马赛克重采样模拟LCD屏幕显示效果,然后利用高斯滤波、JPEG压缩等操作,将生成的图像模拟摩尔纹图像,并将干净图像与摩尔纹图像对齐生成图像对.
CFA Moiré[69]2020128×12811 000个图像对,其中10 000个用
于训练、500个用于验证及500个用于测试的图像对.
干净图像通过从现有的数据库中采集得到,通过马赛克重采样和去马赛过程生成摩尔纹图像,然后经过人工选择做进一步优化.
), ArticleFig(id=1329456190985188328, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=表2, caption=

不同训练数据集的总结对比

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集发布年份图像分辨率(像素)数据规模图像对生成过程
TIP2018[40]2018256×256135 000个图像对,其中90%用于
训练、10%用于验证和测试.
1)图像捕获:将参考图像按照一定的时间间隔在屏幕中显示,用手机拍摄视频,然后提取相应的图像帧;
2)图像对齐:依靠黑色图像边界的角点,计算出投影矩阵,实现图像对齐.
MRBI[67]2021128×128352个训练图像对,340个测试图
像对.
1)图像捕获:利用屏幕截屏和屏幕拍照的方式得到干净图像和摩尔纹图像;
2)图像对齐:基于尺度不变特征找到2个图像中可以匹配的点对,进而得到变换矩阵.
LCD Moiré[68]20191024×102410 200个图像对,其中10 000个用
于训练、100个用于验证及100个用于测试的图像对.
针对干净图像利用马赛克重采样模拟LCD屏幕显示效果,然后利用高斯滤波、JPEG压缩等操作,将生成的图像模拟摩尔纹图像,并将干净图像与摩尔纹图像对齐生成图像对.
CFA Moiré[69]2020128×12811 000个图像对,其中10 000个用
于训练、500个用于验证及500个用于测试的图像对.
干净图像通过从现有的数据库中采集得到,通过马赛克重采样和去马赛过程生成摩尔纹图像,然后经过人工选择做进一步优化.
), ArticleFig(id=1329456191060685801, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Table 3, caption=

Comparative Analysis of the Morié Removal Algorithms Based on Deep Learning

, figureFileSmall=null, figureFileBig=null, tableContent=
算法类别网络模型模型特征及主要优点缺点及不足
CNN空域算法DMCNN[40]利用深度学习去除摩尔纹的最初尝试,实现了图像
多尺度特征提取.
1)忽略了摩尔纹分量在不同尺度间的联系;
2)每一尺度提取的特征不够充分.
MSFE[42]以Unet网络模型为基础,用残差网络模块将输入图像下采样到不同尺度中,每个分支引入特征增强实
现与低分辨率的特征融合.
低分辨率分支中不包含高分辨率特征,导致每一尺度提取的特征不够充分,与DMCNN相比改进不明显.
MopNet[43]利用多尺度网络、边界检测器、模式分类器综合提取摩尔纹的多维特征.1)不同尺度间的信息交换不充分;
2)仅利用边界损失无法处理复杂纹理图像.
HRDN[44]在特征提取的过程中保证了不同尺度间信息的交换和融合.在一个尺度下的特征提取通过一系列的卷积操作,
特征提取不够充分.
MDDM[45]在多分辨率特征提取的同时,利用动态特征编码技术可以高效处理复杂的摩尔纹特征,同时加入色调映射网络,减少输出图像的失真.1)模型的分层数目过多,各个尺度之间缺少
信息交换;
2)不同分辨率的尺度之间卷积数目不合理.
MDDM+[46]基于MDDM存在的不足做了进一步改进,缩减分
支,且每个分支中模块的分配更加合理;加入小波
域损失函数,使重建图像在不同的频段保留更多的
图像细节.
目前存在的绝大多数CNN空域算法中,基本还没有方法尝试对摩尔纹去除进行明确的数学建模.
CNN频域算法WDNet[64]在小波频域内通过准确感知摩尔纹的区域去除摩
尔纹.
忽略了不同频率间摩尔纹分量的联系.
MBCNN[47]把摩尔纹去除的过程看作是纹理恢复和颜色恢复的子过程,在纹理恢复中,通过清除频域中的摩尔纹分量实现摩尔纹的去除,同时改善了输出图像的颜色对
比度.
通过不同数目的扩张卷积得到频谱结果会有不同的尺度特征,不应用一组权值作用于最终的频谱.
MBCNN+[60]在MBCNN基础上做了2个方面的改进:1)用不同的权值矩阵作用于每一次扩张卷积操作后的结果,实现不同频域及不同尺度的摩尔纹去除;2)在计算损失函数时,由原来的ASL变为基于不同扩张尺度的DASL.在目前的监督卷积神经网络中得到较好的效果,但网络结构还有进一步优化的空间.
GAN网络
MR-GAN[49]利用对抗网络去摩尔纹的初步尝试,利用2个生成器和2个鉴别器,同时训练用于在干净图像中加摩尔纹的网络和用于去除摩尔纹的网络,实现在没有成对的训练数据的情况下去除摩尔纹.在网络结构的设计中对摩尔纹的多尺度特征分布考虑不足,而过多依靠复杂的损失函数确保输出图像的质量,图像质量有待进一步提高.
EEUCN[50]创建分别用于添加摩尔纹和去除摩尔纹的网络,将摩尔纹的生成和去除的过程看成是学习摩尔纹的结构和图像强度恢复的过程,并依此构建基于直方图和多尺度特征学习结构.由于用网络添加摩尔纹的图像并不能完全等同于
真实的摩尔纹图像,因此影响了最终的摩尔纹去
除效果.
), ArticleFig(id=1329456191140377578, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=表3, caption=

基于深度学习的摩尔纹去除算法对比分析

, figureFileSmall=null, figureFileBig=null, tableContent=
算法类别网络模型模型特征及主要优点缺点及不足
CNN空域算法DMCNN[40]利用深度学习去除摩尔纹的最初尝试,实现了图像
多尺度特征提取.
1)忽略了摩尔纹分量在不同尺度间的联系;
2)每一尺度提取的特征不够充分.
MSFE[42]以Unet网络模型为基础,用残差网络模块将输入图像下采样到不同尺度中,每个分支引入特征增强实
现与低分辨率的特征融合.
低分辨率分支中不包含高分辨率特征,导致每一尺度提取的特征不够充分,与DMCNN相比改进不明显.
MopNet[43]利用多尺度网络、边界检测器、模式分类器综合提取摩尔纹的多维特征.1)不同尺度间的信息交换不充分;
2)仅利用边界损失无法处理复杂纹理图像.
HRDN[44]在特征提取的过程中保证了不同尺度间信息的交换和融合.在一个尺度下的特征提取通过一系列的卷积操作,
特征提取不够充分.
MDDM[45]在多分辨率特征提取的同时,利用动态特征编码技术可以高效处理复杂的摩尔纹特征,同时加入色调映射网络,减少输出图像的失真.1)模型的分层数目过多,各个尺度之间缺少
信息交换;
2)不同分辨率的尺度之间卷积数目不合理.
MDDM+[46]基于MDDM存在的不足做了进一步改进,缩减分
支,且每个分支中模块的分配更加合理;加入小波
域损失函数,使重建图像在不同的频段保留更多的
图像细节.
目前存在的绝大多数CNN空域算法中,基本还没有方法尝试对摩尔纹去除进行明确的数学建模.
CNN频域算法WDNet[64]在小波频域内通过准确感知摩尔纹的区域去除摩
尔纹.
忽略了不同频率间摩尔纹分量的联系.
MBCNN[47]把摩尔纹去除的过程看作是纹理恢复和颜色恢复的子过程,在纹理恢复中,通过清除频域中的摩尔纹分量实现摩尔纹的去除,同时改善了输出图像的颜色对
比度.
通过不同数目的扩张卷积得到频谱结果会有不同的尺度特征,不应用一组权值作用于最终的频谱.
MBCNN+[60]在MBCNN基础上做了2个方面的改进:1)用不同的权值矩阵作用于每一次扩张卷积操作后的结果,实现不同频域及不同尺度的摩尔纹去除;2)在计算损失函数时,由原来的ASL变为基于不同扩张尺度的DASL.在目前的监督卷积神经网络中得到较好的效果,但网络结构还有进一步优化的空间.
GAN网络
MR-GAN[49]利用对抗网络去摩尔纹的初步尝试,利用2个生成器和2个鉴别器,同时训练用于在干净图像中加摩尔纹的网络和用于去除摩尔纹的网络,实现在没有成对的训练数据的情况下去除摩尔纹.在网络结构的设计中对摩尔纹的多尺度特征分布考虑不足,而过多依靠复杂的损失函数确保输出图像的质量,图像质量有待进一步提高.
EEUCN[50]创建分别用于添加摩尔纹和去除摩尔纹的网络,将摩尔纹的生成和去除的过程看成是学习摩尔纹的结构和图像强度恢复的过程,并依此构建基于直方图和多尺度特征学习结构.由于用网络添加摩尔纹的图像并不能完全等同于
真实的摩尔纹图像,因此影响了最终的摩尔纹去
除效果.
), ArticleFig(id=1329456191484310507, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=EN, label=Table 4, caption=

Quantitative Comparison of Different Network Models

, figureFileSmall=null, figureFileBig=null, tableContent=
检测项目DMCNN[40]MopNet[43]HRDN[44]MBCNN[47]EEUCN[50]
PSNR/dB20.6827.7628.4527.3619.06
SSIM0.8030.8940.9100.9170.628
), ArticleFig(id=1329456191547225068, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, language=CN, label=表4, caption=

不同网络模型的容量对比

, figureFileSmall=null, figureFileBig=null, tableContent=
检测项目DMCNN[40]MopNet[43]HRDN[44]MBCNN[47]EEUCN[50]
PSNR/dB20.6827.7628.4527.3619.06
SSIM0.8030.8940.9100.9170.628
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202220797, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202220797, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202220797, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202220797, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1709308800000, fullTextJson=null, articleText=null, reference=null)
收藏切换
摩尔纹图案自动去除技术综述
亓文法 , 刘宇鑫 , 郭宗明 *
计算机研究与发展 | 图形图像处理 2024,61(3): 728-747
收起
收藏切换
计算机研究与发展 |图形图像处理 2024 , 61 (3) : 728 -747
摩尔纹图案自动去除技术综述
全屏
[Author(id=1329456186623112125, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=qiwenfa@pku.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329456186761524159, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186623112125, language=EN, stringName=Wenfa Qi, firstName=Wenfa, middleName=null, lastName=Qi, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080, bio={"content":"

born in 1978. PhD, assistant professor. Senior member of CCF. His main research interests include information hiding, image processing, machine learning

"}, bioImg=null, bioContent=

born in 1978. PhD, assistant professor. Senior member of CCF. His main research interests include information hiding, image processing, machine learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329456186824438720, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186623112125, language=CN, stringName=亓文法, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1北京大学王选计算机研究所 北京 100080, bio={"img":"ZZXDGMLZQC3DA2r5ZXEr5Q==","content":"

1978年生. 博士,助理研究员. CCF高级会员. 主要研究方向为信息隐藏、图像处理和机器学习

"}, bioImg=ZZXDGMLZQC3DA2r5ZXEr5Q==, bioContent=

1978年生. 博士,助理研究员. CCF高级会员. 主要研究方向为信息隐藏、图像处理和机器学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])]), Author(id=1329456186895741890, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329456186971239364, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186895741890, language=EN, stringName=Yuxin Liu, firstName=Yuxin, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080, bio={"content":"

born in 1985. Master, engineer. Member of CCF. His main research interests include information hiding, image processing, and machine learning

"}, bioImg=null, bioContent=

born in 1985. Master, engineer. Member of CCF. His main research interests include information hiding, image processing, and machine learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329456187055125445, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456186895741890, language=CN, stringName=刘宇鑫, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=1北京大学王选计算机研究所 北京 100080, bio={"img":"h1q2bVeccs+0iOax62itsQ==","content":"

1985年生. 硕士,工程师. CCF会员. 主要研究方向为信息隐藏、图像处理、机器学习

"}, bioImg=h1q2bVeccs+0iOax62itsQ==, bioContent=

1985年生. 硕士,工程师. CCF会员. 主要研究方向为信息隐藏、图像处理、机器学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])]), Author(id=1329456187113845703, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=guozongming@pku.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329456187185148873, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456187113845703, language=EN, stringName=Zongming Guo, firstName=Zongming, middleName=null, lastName=Guo, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080, bio={"content":"

born in 1966. PhD, professor. Distinguished member of CCF. His main research interests include network video technology, image processing, and information security

"}, bioImg=null, bioContent=

born in 1966. PhD, professor. Distinguished member of CCF. His main research interests include network video technology, image processing, and information security

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329456187252257738, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, authorId=1329456187113845703, language=CN, stringName=郭宗明, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=1北京大学王选计算机研究所 北京 100080, bio={"img":"lx625xjoDlabOuL32KTNJQ==","content":"

1966年生. 博士,研究员. CCF杰出会员. 主要研究方向为流媒体技术、图像处理、信息安全

"}, bioImg=lx625xjoDlabOuL32KTNJQ==, bioContent=

1966年生. 博士,研究员. CCF杰出会员. 主要研究方向为流媒体技术、图像处理、信息安全

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329456186501477305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, xref=1, ext=[AuthorCompanyExt(id=1329456186526643130, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080), AuthorCompanyExt(id=1329456186535031739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329456182479139746, companyId=1329456186501477305, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京大学王选计算机研究所 北京 100080)])])]
亓文法 , 刘宇鑫, 郭宗明*
作者信息
  • 1北京大学王选计算机研究所 北京 100080
通讯作者:
郭宗明()
作者简介:

1978年生. 博士,助理研究员. CCF高级会员. 主要研究方向为信息隐藏、图像处理和机器学习

1985年生. 硕士,工程师. CCF会员. 主要研究方向为信息隐藏、图像处理、机器学习

1966年生. 博士,研究员. CCF杰出会员. 主要研究方向为流媒体技术、图像处理、信息安全

Survey of Automatic Removal of Moiré Pattern
Wenfa Qi , Yuxin Liu, Zongming Guo*
Affiliations
  • 1Wangxuan Institute of Computer Technology, Peking University, Beijing 100080
  • born in 1978. PhD, assistant professor. Senior member of CCF. His main research interests include information hiding, image processing, machine learning

    born in 1985. Master, engineer. Member of CCF. His main research interests include information hiding, image processing, and machine learning

    born in 1966. PhD, professor. Distinguished member of CCF. His main research interests include network video technology, image processing, and information security

出版时间: 2024-03-02 doi: 10.7544/issn1000-1239.202220797
文章导航
收藏切换

如今,数码相机和智能手机在人们的生活中扮演着越来越重要的角色,已经成为人们感知世界、记录信息和沟通交流的主要工具. 当使用这些设备拍摄电子屏幕时,显示设备和摄像头传感器网格之间往往会发生混叠,通常导致图片中存在不规则分布的摩尔纹干扰图案,从而严重影响了拍摄图像的视觉质量效果. 因此,摩尔纹图案去除方法研究对于拍摄图像的后期处理具有重要意义. 为此,详细梳理了摩尔纹去除研究的发展脉络,并根据不同的适用场景和技术实现将现有方法分为2类:基于先验知识的摩尔纹去除方法和基于深度学习的摩尔纹去除方法. 鉴于深度学习网络中训练数据集的收集和对齐方式不同,该类方法又分为基于卷积神经网络(CNN)的摩尔纹去除方法和基于生成式对抗网络(GAN)的摩尔纹去除方法. 在此基础上,选择相同的公开数据集,对主流的深度学习方法进行算法实现和性能对比分析,并分别总结了各类方法的优缺点. 最后,对未来的研究方向进行展望.

摩尔纹图案  /  图像恢复  /  多尺度  /  深度学习  /  卷积神经网络  /  生成式对抗网络

Nowadays, digital cameras and smart phones have played an increasingly important role in daily lives, and have become the main tools for people to perceive the world, record information and communicate with each other. When these devices are used to shoot electronic screens, the irregular Moiré patterns in the image are produced due to the overlap of the display devices and digital grids of the camera sensor, which seriously affects the visual quality of the captured images. Therefore, the removal of Moiré pattern is of great significance for the post-processing of captured images. In this paper, recent research about Moiré removal is reviewed in detail, and the existing methods are classified into two categories according to different application scenarios and technical implementations: prior knowledge based Moiré removal and deep learning based methods respectively. According to different training data acquisition and alignment methods, deep learning based Moiré removal techniques can be divided into convolutional neural network (CNN) based methods and generative adversarial network (GAN) based methods. Then, based on the same public dataset, we implement the mainstream deep learning based Moiré removal algorithms, compare and analyze their performance, and summarize the advantages and disadvantages of various methods. Finally, possible directions for future research are discussed.

Moiré pattern  /  image restoration  /  multiscale  /  deep learning  /  convolutional neural network  /  generative adversarial networks
亓文法, 刘宇鑫, 郭宗明. 摩尔纹图案自动去除技术综述. 计算机研究与发展, 2024 , 61 (3) : 728 -747 . DOI: 10.7544/issn1000-1239.202220797
Wenfa Qi, Yuxin Liu, Zongming Guo. Survey of Automatic Removal of Moiré Pattern[J]. Journal of Computer Research and Development, 2024 , 61 (3) : 728 -747 . DOI: 10.7544/issn1000-1239.202220797
在21世纪这个网络信息爆炸的时代,数字图像作为人类认知世界的视觉基础,已经成为信息记录、表达、储存及传播的重要手段. 同时,随着数码相机、平板电脑以及智能手机等移动设备的迅速普及,人们可以轻松地获取各种有意义或有趣的宝贵瞬间作为数字图像的内容. 但是当利用光学镜头设备拍摄电子屏幕(如LED屏、LCD屏等)或者高频重复纹理图案(例如布料、瓷砖等)时,若相机的彩色滤光片阵列(colour filter array, CFA)与拍摄对象条纹之间的空间频率接近,则拍摄图像中会出现不规则的点状、条纹、曲线或者涟漪等形状的干扰图案,即所谓的摩尔纹,如图1所示. 因为从数学上讲,2个具有相近频率的等幅正弦波被叠加,合成信号的振幅将会根据2个频率之间的差异而发生随机变化[1]. 摩尔纹图案的出现严重降低了拍摄图像的视觉质量和美感价值,并且会影响后续的传统图像处理效果,比如图像超分[2]、图像分割[3]、人脸识别[4-5]等. 依据拍摄硬件设备、方向角度以及距离远近的差异,摩尔纹图案存在着形状各异、频谱广泛、纹理随机以及颜色复杂等动态特性. 因此,拍照图像中的摩尔纹图案自动去除面临着巨大的挑战.
理论上讲,摩尔纹去除可以视为传统的图像修复处理,其目的是消除拍照图像中的摩尔纹图案噪声,重建图像的高频细节以及恢复颜色空间信息[6]. 摩尔纹图案去除算法的输入为一幅包含摩尔纹图案的污染图像,输出则为一幅去除摩尔纹图案并进行了颜色空间信息恢复的干净图像. 由于摩尔纹图案和原始图像信号在空域和频域的混合范围都很广,传统的图像去噪[7-8]、去网纹[9-10]和去模糊[11]等图像修复算法不能有效地直接应用于摩尔纹图案去除任务. 比如,图像去噪任务通常假定噪声仅存在高频带中,但摩尔纹的分量不规则地分布于从低频到高频的各个子带中,即表现为在不同图像之间甚至同一图像中的不同区域,摩尔纹图案都会随机占据不同能量的频域子带;图像去网纹技术消除的图像纹理通常是均匀分布的,但是摩尔纹图案的分布疏密不均匀;图像去模糊技术则要求图像噪声的颜色或者强度变化幅度相对一致,而在实际场景中CFA的颜色分布不均衡,使得摩尔纹图案在RGB颜色通道中分别显示不同的强度. 为了抑制摩尔纹图案的产生,最常用的方法是在成像之前添加一些预处理,例如,在相机镜头前面放置一个抗混叠滤波器[12-13],并对CFA的输出图像应用相对复杂的插值算法[14-15]. 由于光学滤波器会导致高频信息丢失,并导致图像过度平滑,这些预处理方法在实践中发挥的能力很有限. 因此,大部分的工作聚焦于图像去摩尔纹的后处理方法研究,尤其是针对拍摄后的屏幕图像.
早期的摩尔纹图案去除研究工作主要是基于摩尔纹生成模型展开的,其中利用了摩尔纹图案的特定先验知识和前提假设. 考虑到常用的Bayer型CFA通常会对采集的颜色通道分量进行插值处理后才能得到拍照后的全彩色图像,因此通过改进图像插值算法可以有效抑制摩尔纹干扰现象的发生[16-17]. 为了充分考虑颜色通道的相关性,更高性能的图像插值算法被提出,如梯度插值方法[18]、自适应插值方法[19]、加权系数插值方法[20]等. 该类方法得到的修复图像能够较好地保留图像的细部边缘特征,但算法实现过于复杂,难以在实际场景中得到广泛应用. 通过对摩尔纹图案的空间结构统计特征和频域能量分布特性进行分析,经过摩尔纹污染的拍摄图像可以被视为摩尔纹图案和背景自然图像的非线性叠加[21-22]. Liu等人[23]和Yang等人[24]分别基于加性模型提出了用于纹理图像摩尔纹和屏摄图像摩尔纹的消除方法,通过图像分解模型实现摩尔纹图案和背景自然图像的区分,并同时保留图像细节以及保证图像锐度. Fang等人[25]将污染屏摄图像描述为潜在层和摩尔纹图案层,结合潜在层的分段常数特性,提出了一种凸模型来解决摩尔纹消除问题. Sasada等人[26]和Sidorov等人[27]基于摩尔纹图案具有特定形状(如条纹、点状或单色)的假设来进行检测,并将摩尔纹图案与背景层相分离,实现摩尔纹图案的去除和背景颜色通道信息的恢复. 另外,由于摩尔纹图案可以视为一种高频噪声,很多学者通过对图像频域能量信号进行分析,并应用相应滤波器将摩尔纹图案作为特定频率信号去除. 常见的滤波器包括中值滤波器[28-29]、高斯陷波滤波器[28]、DoG(difference of Gaussians)[30]、非线性滤波器[31-32]、自适应滤波器[33]等. 但是该类方法使用到的滤波器通常是经过人工设计的,去除摩尔纹图案后的图像会出现局部过平滑现象,部分图像细节缺失. 另外,当背景自然图像中的感兴趣区域(region of interest, ROI)跟摩尔纹图案具有同样的高频特性时,基于滤波的摩尔纹图案消除方法就会失效. 总之,基于传统信号处理的方法通常依赖于摩尔纹图案生成的先验知识,计算复杂度高,而真实世界中的摩尔纹图案往往具有不同的形状和频域特征,此时该类方法的处理效率较低.
近年来,基于卷积神经网络(convolutional neural network, CNN)的学习方法得到了长足发展,成为计算机视觉和图像处理领域的一场革命. CNN在图像分类和图像识别方面取得成功后,同样在低水平视觉和图像处理任务中也被证明是非常有效的,包括图像超分辨率[34-35]、去马赛克[36]、图像去噪[37]和图像重建[38]等. Abraham[39]将CNN方法引入摩尔纹图案处理领域,提出了一种利用小波分解和多输入深度卷积神经网络模型来检测计算机屏幕拍摄图像(屏摄图像)中的高频摩尔纹噪声图案. 由于摩尔纹图案跨越很宽的频域范围,Sun等人[40]利用一种多分辨率全卷积神经网络(deep multiresolution fully convolutional neural network, DMCNN)自动去除屏摄图像中的摩尔纹图案,该网络在计算如何消除每个频带内的摩尔纹伪影之前,对输入图像进行非线性多分辨率分析. 之后,更多基于空域的多尺度CNN模型[1,41-46]被相继提出,屏摄图像中的摩尔纹图案去除性能得到显著提高. 除了考虑摩尔纹图案的空间特征外,Zheng等人[47]提出了可学习的多尺度带通滤波器,以处理频域中摩尔纹图案的多样性. 此外,空间域和频域变换也被用来探究摩尔纹图案的互补特性,高效地实现自然图像和纹理的恢复. 然而,文献[1,41-47]的方法基于监督学习需要大量的可以成对的干净图像−摩尔纹图像进行训练,其模型性能在很大程度上取决于训练对的特性. 为了解决这一限制,最近研究了使用生成性对抗网络(generative adversarial network, GAN)的无监督学习算法[48-50],由摩尔纹图案生成网络和摩尔纹图案去除网络组成. 生成网络负责生成摩尔纹图案以构造摩尔纹和干净图像的伪配对集,然后使用生成的伪配对数据集以有监督的方式对摩尔纹图案去除网络进行训练,以有效地去除摩尔纹图案.
综上所述,摩尔纹图案去除算法主要包括基于传统信号处理方法和基于深度学习方法. 鉴于摩尔纹图案的频率分布复杂、颜色通道幅度的不平衡以及外观属性不同等特点,深度学习方法在摩尔纹图案去除方面的性能更加出众. 除了如何构造有效的摩尔纹图案去除网络结构外,大规模图像对的基准数据库构建对于摩尔纹去除算法的研究和评估也具有重要意义.
本文的主要贡献包括3个方面:
1)系统梳理了摩尔纹图案去除方法的研究脉络,并进行合理的分类归纳总结;
2)基于相同的公开基准数据库,选择主流的基于深度学习方法进行算法实现和性能对比分析,并总结了相应方法的优缺点;
3)对目前的摩尔纹图案去除算法的研究现状进行总结,并对未来的研究方向进行展望.
当使用智能手机或数码相机等设备拍摄屏幕图像时,由于显示设备像素和摄像头传感器网格之间的混叠,导致传感器对场景图像产生欠采样,因而产生频率干涉的摩尔纹图案. 为此,数码相机开发商提出了一种光学低通滤波器(optical low pass filter, OLPF)方案[12],主要是使用2个透镜将光信号折射至2个方向,从而降低信号源的频率以避免混叠. 在此基础上,Schoberl等人[13]计算求得了一组可以达到最小信号频率混叠和最好图像分辨率的滤波器参数. 除了 OLPF 这种前置滤波方法以外,“后处理”的滤波方法也在专业领域得到了应用.
Wei等人[28]提出了一种从扫描透射X射线显微镜图像中滤除摩尔纹图案噪声的后处理方法. 该方法包括使用局部中值滤波器半自动检测傅里叶振幅谱中的谱峰,以及使用高斯陷波滤波器消除谱噪声峰值. 文献[30]通过使用DoG对傅里叶图像进行滤波,分析图像中的摩尔纹图案来识别人脸欺骗. 尽管高频滤波器可以检测摩尔纹图案,但它无法将其与其他感兴趣的高频对象区分开,这可能会导致误报. Sidorov等人[31]通过研究摩尔纹变形模型,基于图像傅里叶频谱幅值的阈值化设计了非线性滤波器,解决通过硬件设备在电影到视频的数字转换过程中出现的摩尔纹干扰问题,以减少由于摩尔纹图案的非平稳性而可能出现的振铃伪影的影响. 基于扫描半色调图像模型,Sun等人[33]提出了一种基于自适应滤波的去网纹方法,从扫描图像中恢复出高质量的连续色调图像. 首先,采用基于图像冗余的去噪算法来降低打印噪声和衰减失真;然后,使用扫描图像的屏幕频率和局部梯度特征进行自适应滤波;最后,使用边缘保留滤波器进一步增强边缘的清晰度,以恢复高质量的连续色调图像. 文献[28-33]的方法均将摩尔纹当成了一种高频噪声,通过去除噪声的方式达到摩尔纹去除的目的. 然而,与噪声不同的是,摩尔纹图案广泛分布在图像的各个频带,传统的基于图像滤波的方法并不能完全解决摩尔纹去除的问题.
拍照设备传感器表面覆盖了CFA,传感器则具有感知光线不同色彩强度的能力,其中每个滤光片仅允许RGB 3原色中的其中一种颜色通过,即CFA仅仅采集到了原始图像1/3的信息量. 为了重建原始图像,需要利用插值的方法恢复另外2个颜色,这种方法也称为去马赛克. 通过利用颜色通道之间的相关性,邱菊[16]提出了一种基于图像色差的插值算法进行图像摩尔纹的去除,并通过将向量的概念引入到插值算法中实现了RGB通道的3维插值,恢复后的图像更加平滑、更接近实际值. 在文献[16]的基础上,邱香香[17]则利用自适应方法选择关联度高的颜色分量来协助判断通道插值方向,图像插值结果更为合理. Hamilton[19]通过计算图像亮度及色度的相近程度确定针对图像的插值方法;Kimmel[20]提出了以不同权值为基础的插值方法,权值与邻域内边缘的信息密切相关,每个像素缺失的颜色分量由邻域内像素的颜色分量按照不同的权值计算得到. 此外,Hibbard[18]利用图像在水平和竖直方向上的亮度和色度差计算梯度,然后将得到的梯度值与事先设定的阈值进行比较来确定针对亮度和色度进行插值的方向,最后根据方向做插值运算. 这类方法得到的恢复图像能够较好地保留图像前景部分的边缘细节特征,但算法实现过于复杂,尤其难以在移动设备中实现.
另外一类常用的“后处理”即为基于图像分解实现摩尔纹图案的消除. Liu等人[23]提出了一种低秩稀疏矩阵分解模型,实现织物图像摩尔纹的消除. 通过对纹理分量和摩尔纹分量分别进行空域和频域分析,发现摩尔纹分量分布集中,几乎不与纹理分量的能量混合,于是对纹理分量添加低秩先验约束,对摩尔纹分量添加稀疏先验约束以及在其频域分布内添加位置约束,从而区分摩尔纹分量和纹理分量. 后来,Yang 等人[51]改进文献[23]所提算法,从数码相机拍摄织物图像出现摩尔纹的成像原理入手,发现伪彩色波纹状的摩尔纹主要存在于R,B这2个通道,G通道的摩尔纹分量较少,于是将文献[23]中的方法应用于G通道的图像信息,再通过RGB 三通道间的相关性,借助已经恢复的G通道图像,并应用导向滤波算法恢复R,B通道图像. Ok 等人[52]提出了一种纸质支票上的摩尔纹消除方法,通过前景提取、摩尔纹检测和摩尔纹消除等操作提高图像质量. Yang等人[24]以摩尔纹和背景图像是加性关系为前提,并结合屏摄摩尔纹的3通道成像差异和结构相似性,提出了联合小波域导向滤波和基于高斯混合模型(Gaussian mixed model, GMM)的多相图层分解模型以消除屏摄摩尔纹图案.
屏摄图像$ I $可以看作是背景图层$ B $、摩尔纹图层 $ M $以及方差为$ {\sigma ^2} $的高斯噪声$ n $的叠加,假设有加性模型[17]为:
$ I = B + M + n, $
该模型引入了2个基于 GMM的图像先验,分别对背景图层和摩尔纹图层进行规则化. 于是该问题转化为最小化问题:
$ \underset{B,M}{\mathrm{min}}\frac{1}{{\sigma }^{2}}{\Vert I-B-M\Vert }_{2}^{2}+\phi (B)+\psi (M)\text{,} $
其中$ \left\| {I - B - M} \right\|_2^2 $是保真项,$ \phi (B) $和$ \psi (M) $分别定义了先验项和约束项,完成对$ \left\| {I - B - M} \right\|_2^2 $的规则化. 通过上述基于GMM的图层分解模型最终获得无摩尔纹的屏摄图像$ B $.
在实际情况下,摩尔纹分量与纹理分量在频域存在能量混叠,因此基于图像分解方法容易将纹理信息误判为摩尔纹成分,纹理恢复图像中存在轻微的振铃效应. 另外,由于导向滤波算法本身固有的缺陷,容易导致恢复图像的局部出现光晕现象.
基于传统信号处理的方法探究了摩尔纹图案的成因以及相关特性的先验知识,可以用于特定场景下的摩尔纹图案自动去除,各类方法的具体对比分析如表1所示. 然而鉴于摩尔纹图像本身具有分布不规则的特性,传统方法很难将分布在不同频段的摩尔纹分量去除干净,因此该类方法的摩尔纹图案自动去除性能还需要进一步提升.
基于CNN的摩尔纹去除方法主要包括基于图像空间域的分析模型和基于图像频域的分析模型,以下分别选取具有代表性的网络结构模型进行介绍.
由于摩尔纹图案跨越很宽的频域范围,Sun等人[40]最先尝试利用CNN去除摩尔纹图案,提出了一种基于DMCNN的摩尔纹去除方法,其网络结构如图2所示. 该网络共有5个平行的分支,最顶层的分支是在输入图像原始的分辨率下处理,下层的分支通过由上层分支做下采样得到. 每个分支分别做各自的卷积操作,最后进行反卷积运算将各个分支的结果上采样到原始图像的分辨率后相加,从而得到输出的恢复图像. Liu等人[41]提出了由1个粗尺度网络和1个细尺度网络组成的2段式摩尔纹去除方法DCNN(deep CNN),在粗尺度网络上对输入图像做降采样,利用堆叠的残差块去除摩尔纹,然后利用细尺度网络上采样至原始的分辨率.
以DMCNN为基础,为了利用不同尺度特征映射之间的关系,Gao等人[42]提出了一种基于深度学习的多尺度特征增强(multi-scale feature enhancing, MSFE)摩尔纹去除网络模型,并利用多尺度的结构从多个分辨率上提取与摩尔纹无关的上下文. 该方法以U-net网络模型[53]为基础,首先利用残差网络模块,将输入图像下采样到4个不同的尺度中,每个分支在特征增强处理过程中,实现将所在分支的特征与所有低于所在分支分辨率的特征相融合,最后将每个分支利用残差模块做上采样处理,不同分支的采样结果融合后得到最终的输出图像.
针对摩尔纹图案的频率分布复杂、颜色通道幅度的不平衡以及外观属性不同等特点,He等人[43]提出了MopNet(Moiré pattern removal neural network)神经网络模型用于摩尔纹图案去除,具体包括:
1) 多尺度特征聚合. 首先提取图像的多尺度特征,然后通过级联[54]和SE (squeeze-and-excitation) 块[55]进行特征融合. SE块通过计算每个通道的归一化权重来重新加权特征图,所产生的聚合特征可以用于如下数学方法表示:
$ {\xi }_{\text{m}}=SE(Cat[N{U}_{1}({F}_{1}),N{U}_{2}({F}_{2}),\cdots ,N{U}_{n}({F}_{n})])\text{,} $
其中$ SE $为SE块操作,$ Cat $表示串联,$ {F_i} $表示从多尺度提取器$ {g_{\text{m}}} $获得的不同频带的特征图,$ N{U_i} $是用于将特征图转换为相同空间大小的非线性上采样.
2) 信道方向目标边缘预测器. 利用网络来预测无摩尔纹目标图像的通道方向边缘映射$ {E_{\text{p}}} $,并使用Sobel算子将每个颜色通道的$ {E_{{\text{sr}}}} $,$ {E_{{\text{sg}}}} $,$ {E_{{\text{sb}}}} $的单独边缘映射用于增强源图像$ {I_{\text{s}}} $,边缘预测器为
$ {E}_{\text{p}}=Cat({E}_{\text{pr}},{E}_{\text{pg}}, {E}_{\text{pb}})={g}_{\text{e}}({I}_{\text{s}},{E}_{\text{sr}},{E}_{\text{sg}},{E}_{\text{sb}}), $
其中$ {g_{\text{e}}} $为信道边缘预测器,$ {E_{{\text{pr}}}} $,$ {E_{{\text{pg}}}} $,$ {E_{{\text{pb}}}} $表示R,G,B信道的预测边缘图.
3) 基于属性的摩尔纹模式分类器. 摩尔纹图案的精确描述可以更好地指导学习过程,因此,使用多标签分类器$ C $来描述摩尔纹图案的频域$ {C_0} $、颜色$ {C_1} $和形状$ {C_2} $这3个外观属性,并通过连接3个上采样标签映射获取属性信息:
$ C=Cat[u({C}_{0}),u({C}_{1}),u({C}_{2})]\text{,} $
然后将这种模式属性的预测输入到目标输出的推理中,为摩尔纹图案的外观提供辅助指导.
MopNet的所有核心组件都是专为摩尔纹图案的独特特性而设计的,包括使用多尺度特征聚合以解决复杂频率问题、使用通道方向的目标边缘预测器以检测颜色通道之间的不平衡幅度,以及使用属性感知分类器以表征不同外观,并更好地建模摩尔纹图案.
为了克服MopNet以及MSFE模型存在的多尺度信息交换和融合的不足,Yang等人[44]提出了一种基于高分辨率的摩尔纹去除网络(high-resolution deMoiré network,HRDN),以充分探索不同分辨率特征映射之间的关系. HRDN由3个主要部分组成:并行高分辨率网络、连续信息交换模块和最终特征融合层,其结构如图3所示. HRDN网络是一种带有残差块的并行多尺度架构,它可以在处理较低分辨率的同时保持全分辨率,并处理不同尺度的不同频率. 为了充分利用不同特征图之间的关系,HRDN使用信息交换模块和最终特征融合层在整个网络中不断地交换信息,以充分融合从低层到高层的特征,反之亦然. 通常情况下像素损失函数会导致过度平滑,从而减少细节和边缘纹理. 为此,HRDN通过定义边缘增强损失函数来保留边缘细节,L可以表示为
$ L=\alpha \times CLoss+(1-\alpha )\times SLoss\text{,} $
其中$ CLoss $表示L1 Charbonnier损失,$ SLoss $表示L1 Sobel损失,分别定义为:
$ \begin{array}{l}CLoss=\dfrac{1}{N}{\displaystyle \sum _{i=1}^{N}\sqrt{{(\hat{Y}_i-Y_i)}^{2}-{\varepsilon }^{2}}}\text{,}\\ SLoss=\dfrac{1}{N}{\displaystyle \sum _{i=1}^{N}\sqrt{{(Sobel(\hat{Y}_i)-Sobel(Y_i))}^{2}}}\text{,}\end{array} $
其中N代表批量大小,$ \hat Y $和$ Y $分别表示HRDN的输出去摩尔纹图像和原始背景干净图像.
HRDN模型在图像分解和图像融合中都保留了摩尔纹分量在不同尺度下的综合特征,摩尔纹去除效果有了明显改进. 但是在每一个尺度中,仅通过一系列卷积操作,无法提取完整的特征信息. 为此,Cheng等人[45]提出了一种基于多尺度动态特征编码的摩尔纹去除网络(multi-scale convolutional network with dynamic feature encoding for image deMoiréing, MDDM). 该算法首先利用降采样的方式将输入图像分成6层尺度,分辨率最高的一层仅做卷积运算,将基于通道的动态特征编码的残差模块作用于从第2层到第6层尺度的分支,以实现在复杂的纹理背景下区分干净图像与被摩尔纹污染的图像数据. MDDM通过动态特征编码来增强模型处理动态复杂纹理的能力,并将全局残差学习方法应用于网络中每个分辨率的分支;它使用残差块在每个特征级别和频带上模拟干净图像和摩尔纹图像之间的差异,即每个分支上的摩尔纹图案. 在动态特征编码的过程中,利用自适应实例归一化(adaptive instance normalization,AdaIN)方法[56]将额外的旁路分支引入至每个主干尺度分支中,以对不同空间分辨率的图像特征进行编码. 具体地,在AdaIN中,首先计算特征图的均值和方差:
$ {\hat{\mu }}_{i}=\dfrac{1}{HW}{\displaystyle \sum _{j=1}^{H}{\displaystyle \sum _{k=1}^{W}{x}_{ijk}^{\text{enc}}}}\text{,} $
$ {\hat{\sigma }}^{2}_i=\frac{1}{HW}{\displaystyle \sum _{j=1}^{H}{\displaystyle \sum _{k=1}^{W}{({x}_{ijk}^{\text{enc}}-{\hat{\mu }}_{i})}^{2}}}\text{,} $
其中H和W分别表示特征图的高度和宽度,$ {\hat \mu _i} $和$ \hat \sigma _i^2 $是动态特征编码分支中第i个编码层的特征$ {x^{{\text{enc}}}} $的均值和方差. 在计算摩尔纹图案的统计值后,MDDM使用这些值通过AdaIN动态调整主干分辨率分支的参数:
$ {x}_{i+1}=\frac{{x}_{i}-{\mu }_{i}}{\sqrt{{\sigma }_{i}^{2}}+\varepsilon }\sqrt{{\hat{\sigma }_i}^{2}}+{\hat{\mu }_i}^{2}\text{,} $
其中$ {\mu _i} $和$ \sigma _i^2 $表示来自主干分支的统计信息,$ {x_i} $表示来自主干分支中第i个残差块的特征图.
在MDDM的基础上, Cheng等人[46]又进一步提出了改进的MDDM+模型. 整个网络由原来的6个分支减少为3个分支,每个分支中基于通道的动态特征编码的残差模块的数目保持一致,这样可以大幅度减少存储模型所需的参数数目,同时也改进了模型在不同分辨率的尺度间参数不平衡的现象. 另外,考虑到摩尔纹的频域分布特征,引入了基于小波分解的特殊损失函数,即分别对网络的输出图像$ \hat I $和ground truth图像$ I $做小波变换,小波损失表示为
$ WLos{s}=\frac{1}{N}{\displaystyle \sum _{i=1}^{N}\sqrt{({\displaystyle \sum _{w=1}^{4}W_w(\hat{I}_i)-{\displaystyle \sum _{w=1}^{4}W_w(I_i){)}^{2}+{\varepsilon }^{2}}\text{,}}}} $
其中$ N $表示一个批量大小,$ \varepsilon $为一个常数,$ \varepsilon $=E-6,$ W $即小波分解操作.
综上,除了将单幅图像作为输入外,Liu等人[1]提出了一种基于多帧及多尺度的摩尔纹图案去除(multi-frame and multi-scale for image deMoiréing, MMDM)网络. MMDM使用多幅图像作为输入,多尺度特征编码模块用于低频信息增强. MMDM有3个关键模块:新设计的多帧空间变换网络(multiframe spatial transformer network, M-STN)、多尺度特征编码模块(multiscale feature encoding module, MSFE)和增强非对称卷积块(enhanced asymmetric convolution block, EACB). 其中M-STN自动对齐相同场景下的多帧输入图像,MSFE用于多频率信息的提取,最终EACB用于图像的重建. 在NTIRE2020(New Trends in Image Restoration and Enhancement 2020)挑战赛中,本方法在摩尔纹去除赛道上获得第2名.
除了针对摩尔纹分量在空域上的分布特征展开研究外,一些学者还尝试提取摩尔纹分量在图像频域内的特征并进行摩尔纹去除[47,57-61]. Zheng等人[47]认为摩尔纹去除的过程可通过式(12)表示
$ {I}_{\text{clean}}={\psi }^{-1}({I}_{\text{moire}}-{N}_{\text{moire}})\text{,} $
其中$ {I_{{\text{moire}}}} $与$ {I_{{\text{clean}}}} $分别为包含摩尔纹的污染图像和去除摩尔纹后的干净图像,$ {N_{{\text{moire}}}} $为污染图像中的摩尔纹分量,$ {\psi ^{ - 1}} $是$ {\psi ^{}} $的反函数,即将污染图像中的颜色色调恢复为原始干净图像色调. 基于此理论提出的基于多尺度带通CNN(multi-scale bandpass CNN, MBCNN)网络分别解决了2个子问题:1)纹理恢复问题. 在摩尔纹纹理去除模块(Moiré texture removal block, MTRB)中集成了一种可以学习的带通滤波器,以学习去除摩尔纹前后的频率特征. 2)颜色恢复问题. 首先利用全局色调映射模块(global tone mapping block, GTMB)校正图像的全局色彩,然后利用局部色调映射模块(local tone mapping block, LTMB)对局部色调逐个进行像素级别的微调. 模型首先将输入图像分成4个大小相等的子图像,网络结构由3个分支构成,每个分支依次执行特定比例的摩尔纹去除和色调映射,最终输出上采样后的图像,并与上层分支融合. 在前2个分支中,将当前分支的特征与下一层分支的输出结果相融合,然后再进行一系列的色调处理和摩尔纹去除,以消除因缩放操作引起的纹理和颜色错误.
摩尔纹分量$ {N_{{\text{moire}}}} $表示为
$ {N}_{\text{moire}}={D}^{-1}({\displaystyle \sum _{i}{\displaystyle \sum _{j}F{S}_{{f}_{ij}}^{{S}_{i}})}}\text{,} $
其中$ FS_{{f_{ij}}}^{{S_i}} $为摩尔纹分量在尺度为$ {S_i} $、频率为$ {f_{ij}} $时的频谱. DCT域的变换是线性的,可通过一系列卷积层来建模. 鉴于摩尔纹分量通常会分布在不同的频率中,为此,在摩尔纹去除模块MTRB中,利用可学习的带通滤波器(learnable bandpass filter, LBF)为每一个频率引入一个可以学习的权重. MTRB的网络模型如图4(a)所示,输入为$ x_{{\text{in}}}^{{\text{MTRB}}} $,其中紧致模块(dense block)用于提取特征,该模块由k个紧密连接(densely connected)[62]的扩张卷积层(dilated convolution)[63]与ReLU层组成,输入数据经过紧致模块得到紧致特征$ {F_{{\text{deep}}}} $,将一个3×3的卷积层作用于$ {F_{{\text{deep}}}} $得到了图像在频域中的分量$ \zeta $. 去摩尔纹的过程可以看作是利用一组可学习的权值矩阵U作用于$ \zeta $,得到干净图像的频域分量,最后经过DCT逆变换得到输出结果. 为了避免逆变换后的结果超出实际图像的颜色值范围,利用特征幅度层(feature scale layer, FSL)线性约束输出结果.
在损失函数中引入了改进的Sobel算子损失(advanced sobel loss, ASL),定义为
$ ASL(\hat{Z},Z)=\frac{1}{N}{\displaystyle \sum_{i=1}^N \left|Sobe{l}^{*}(\hat{Z}_i)-Sobe{l}^{*}(Z_i)\right|}\text{,} $
其中$ \hat Z_i $表示网络输出的结果,$ Z_i $表示ground truth,相较于传统的$ Sobel $运算,改进的$ Sobe{l^*} $除了提取水平和竖直方向的结构特征,还提取图像在对角方向的结构特征,信息提取更加丰富.
在提取摩尔纹分量特征时,MBCNN中的每个摩尔纹去除模块都用一组相同的权值作用于经过一系列卷积运算得到的频谱中. 鉴于摩尔纹的多尺度频域分布特征,Zheng等人[60]在原有的MBCNN基础上做了2方面改进,这里称为MBCNN+:
1) 将摩尔纹去除模块MTRB中带通滤波器LBF变为基于不同块大小的带通滤波器(multi-block-size LBFs),即由原来经过k个扩张卷积运算后的结果进行最终的一次频域反变换,转变为每一次扩张卷积操作后都进行一次频域反变换,以达到学习摩尔纹分量不同尺度、不同频域特征的目的,最后将k个不同尺度结果累加后送入FSL,改进后的MTRB+结构如图4(b)所示.
2) 在计算损失函数时,由原来的ASL运算变为基于不同扩张尺度的ASL,即DASL:
$ DASL={\displaystyle \sum _{i=1}^{n}ASL|{}_{{d}_{i}}}\text{,} $
其中$ ASL\left| {_{{d_i}}} \right. $是扩张尺度在$ {d_i} $下的ASL.
Liu等人[61]设计了一种基于小波域的双分支网络(wavelet-based dual-branch network,WDNet). Liu等人认为在频域中更容易去除摩尔纹图案,首先使用小波变换将包含摩尔纹的输入图像分解为不同的频带;经过小波变换后,摩尔纹在某些小波子带中更加明显,在这些子带中,摩尔纹更容易去除. 该模型具有密集分支和扩张分支的双分支网络,分别负责恢复近距离和恢复远距离信息. 同时,WDNet还设计了一种空间注意力机制,称为密集分支中的方向感知模块,以突出具有摩尔纹图案的区域.
流行的基于监督学习的方法需要大量的成对训练图像,这在实际应用中难以获取和对齐. 为此,基于GAN的非监督学习方法被应用于摩尔纹去除[48-50].Yue等人[49]提出了一种无监督的生成式对抗性摩尔纹去除网络(unsupervised generative adversarial network for Moiré removal,MR-GAN),这是基于无监督学习的摩尔纹去除的首次尝试,为使用未配对的训练数据进行无监督摩尔纹去除开辟了一条途径.
具体地,在进入网络训练之前,首先根据图像不对称的径向梯度分布[64]去除摩尔纹图像中存在的晕角,校正后的图像亮度比原始屏幕拍摄图像更加均匀. 在MR-GAN网络中存在2个生成器,即摩尔纹去除生成器和摩尔纹重构生成器. 摩尔纹去除生成器$ {G_{{\text{m}} \to {\text{c}}}} $是将一幅摩尔纹图像$ {I_{\text{m}}} $转化成相应的干净图像$ {\tilde I_{\text{c}}} $;摩尔纹重构生成器$ {G_{{\text{c}} \to {\text{m}}}} $是将一幅干净的图像转化为摩尔纹图像. $ {G_{{\text{m}} \to {\text{c}}}} $和$ {G_{{\text{c}} \to {\text{m}}}} $形成一个循环,并受到循环一致损失的约束,以在没有训练对的情况下同时训练生成器$ {G_{{\text{m}} \to {\text{c}}}} $和$ {G_{{\text{c}} \to {\text{m}}}} $. 这2个生成器具有相同的网络结构,核心结构均是9个残差模块的堆叠. 同时,网络同样存在2个结构相同的鉴别器:$ D_{{\text{m,c}}}^{\text{l}}\underline {\underline \Delta } \left\{ {D_{\text{m}}^{\text{l}},D_{\text{c}}^{\text{l}}} \right\} $在大尺度上鉴别摩尔纹图像与干净图像,$ D_{{\text{m,c}}}^{\text{s}}\underline{\underline \Delta } \left\{ {D_{\text{m}}^{\text{s}},D_{\text{c}}^{\text{s}}} \right\} $在小尺度上鉴别摩尔纹图像与干净图像,核心结构是数个卷积层+ NL(non-local)层 [65]+Leaky ReLUs的堆叠,最后再加1层卷积. 为了训练MR-GAN,集成了许多损失函数来学习真实的无摩尔纹图像生成器和有效的鉴别器,具体包括:
1) 生成器的损失函数. 为了在没有真实监督的情况下学习有效的生成器,该方法首先提出了一组自我监督的损失函数$ {\vartheta _{{\text{self}}}} $,包括在像素级别 $ {\vartheta _{{\text{pcyc}}}} $和特征级别 $ {\vartheta _{{\text{fcyc}}}} $的循环一致损失、身份损失$ {\vartheta _{{\text{idt}}}} $、余弦相似度损失$ {\vartheta _{{\text{cos}}}} $以及内容泄漏损失$ {\vartheta _{{\text{cl}}}} $. 因此,自监督损失表示为
$ {\vartheta }_{\text{self}}={\lambda }_{1}{\vartheta }_{\text{pcyc}}+{\lambda }_{2}{\vartheta }_{\text{fcyc}}+{\lambda }_{3}{\vartheta }_{\text{idt}}+{\lambda }_{4}{\vartheta }_{\text{cos}}+{\lambda }_{5}{\vartheta }_{\text{cl}}\text{,} $
其中$ {\lambda _1} $,$ {\lambda _2} $,$ {\lambda _3} $,$ {\lambda _4} $,$ {\lambda _5} $分别为加权参数. 另外,与传统的对抗性损失不同,该方法提出了一种双尺度对抗性损失分别对应于MR-GAN 中的2个鉴别器组. 为了确保网络在训练中稳定,利用 LSGAN 损失函数[66]定义大尺度特征的对抗损失$ \vartheta _{\text{G}}^{\text{l}} $:
$ \begin{split}{\vartheta }_{\text{G}}^{\text{l}}=& {E}_{{I}_{\text{m}}\sim{P}_{\text{data}}(\mu )}{(}{{D}_{\text{c}}^{\text{l}}}(G_{{\mathrm{m}}\to {\mathrm{c}}}(I_{\mathrm{m}}))-1)^2+\\& {E}_{{I}_{{\text{c}}\sim}{P}_{\text{data}}(\ell )}{(}{{D}_{\text{m}}^{\text{l}}}(G_{{\mathrm{c}}\to {\mathrm{m}}}(I_{\mathrm{c}}))-1)^2,\end{split} $
小尺度特征的对抗损失$ \vartheta _{\text{G}}^{\text{s}} $具有与${\vartheta }_{\text{G}}^{\text{l}} $相同的结构,其中将鉴别器$ ({D}_{\text{c}}^{\text{l}},{D}_{\text{m}}^{\text{l}}) $替换为$ ({D}_{\text{c}}^{\text{s}},{D}_{\text{m}}^{\text{s}}) $. 综上所有损失函数,训练生成器的完整损失函数表示为
$ {\vartheta }_{\text{G}}={\vartheta }_{\text{self}}+{\vartheta }_{\text{G}}^{\text{l}}+{\vartheta }_{\text{G}}^{\text{s}}\text{,} $
其中$ {\vartheta _{{\text{self}}}} $确保生成的结果包含足够的原始信息,$ \vartheta _{\text{G}}^{\text{l}} $和$ \vartheta _{\text{G}}^{\text{s}} $指导生成器得到预期结果.
2) 鉴别器的损失函数. 分别用于训练判别器$ D_{\text{c}}^{\text{l}} $和$ D_{\text{m}}^{\text{l}} $的损失函数$ {\vartheta _{D_{\text{c}}^{\text{l}}}} $和$ {\vartheta _{D_{\text{m}}^{\text{l}}}} $定义为
$ \begin{split}{\vartheta }_{{D}_{\text{c}}^{\text{l}}}=& \frac{1}{2}{E}_{{I}_{\text{m}}\sim{P}_{\text{data}}(\mu )}{(}{{D}_{\text{c}}^{\text{l}}}(G_{{\mathrm{m}}\to {\mathrm{c}}}(I_{\mathrm{m}})))^2+\\& \frac{1}{2}{E}_{{I}_{\text{c}}\sim{P}_{\text{data}}(\ell )}{(}{{D}_{\text{c}}^{\text{l}}}(I_{{\mathrm{c}}})-1)^2,\\ {\vartheta }_{{D}_{\text{m}}^{\text{l}}}=&\frac{1}{2}{E}_{{I}_{\text{c}}\sim{P}_{\text{data}}(\ell )}{(}{{D}_{\text{m}}^{\text{l}}}(G_{{\mathrm{c}}\to {\mathrm{m}}}(I_{\mathrm{c}})))^2+\\& \frac{1}{2}{E}_{{I}_{\text{m}}\sim{P}_{\text{data}}(\ell )}{(}{{D}_{\text{m}}^{\text{l}}}(I_{\mathrm{m}})-1)^2.\end{split} $
小尺度特征鉴别器的损失函数$ {\vartheta }_{{D}_{\text{c}}^{\text{s}}},{\vartheta }_{{D}_{\text{m}}^{\text{s}}}$定义与式(19)类似. 通过最小化$ {\vartheta _{D_{\text{c}}^{\text{l}}}} $,$ {\vartheta _{D_{\text{m}}^{\text{l}}}} $,$ {\vartheta _{D_{\text{c}}^{\text{s}}}} $,$ {\vartheta _{D_{\text{m}}^{\text{s}}}} $,MR-GAN网络鉴别器可以在大尺度和小尺度上区分生成的无摩尔纹图像和真实的干净图像. 该方法不仅可以去除摩尔纹图案,还尽量避免了恢复图像在亮度和对比度上可能存在的失真.
Park等人[50]提出了一种基于端到端的非对称循环网络(an end-to-end unpaired cyclic network,EEUCN),该方法使用未配对的摩尔纹图像和干净的图像数据集. 令$ {{\chi }} $和$ {{\gamma }} $分别表示摩尔纹图像集和干净图像集,$ X \in {{\chi }} $和$ Y \in {{\gamma }} $分别表示摩尔纹图像和干净图像,未配对图像去摩尔纹的目标就是学习从摩尔纹图像到干净图像之间的映射. 如图5所示,EEUCN网络由2种类型的GAN组成:图5(a)所示的摩尔纹生成网络和图5(b)所示的摩尔纹去除网络. 摩尔纹生成网络主要学习映射$ {G}_{\text{M}}:\gamma \to \chi $,通过添加摩尔纹伪影来降低干净图像质量;摩尔纹去除网络则是学习映射$ {G_{\text{D}}}:{{\chi }} \to {{\gamma }} $,以从摩尔纹图像中去除摩尔纹伪影干扰. 摩尔纹生成问题可以分为像素强度退化和摩尔纹图案生成2个子问题,相应地,屏摄摩尔纹图像生成器模型$ {G_{\text{M}}} $可以表示为
$ X=\alpha Y+m\text{,} $
其中$ m $表示摩尔纹图案,$ \alpha $用于控制图像对比度降低的幅度,$\alpha\in (0,1] $. 给定一张干净的图像 Y,强度退化模块估计全局强度退化参数α,而摩尔纹生成模块生成摩尔纹图案 m.
摩尔纹生成网络构造完伪配对集$ \{{{\gamma}} ,{G}_{\text{M}}({{\gamma}} )\} $后,EEUCN利用学习的生成器$ {G_{\text{M}}} $以有监督方式训练摩尔纹去除网络,并进行摩尔纹图案去除,如图5(b)所示. 相应地,摩尔纹去除模型$ {G_{\text{D}}} $可以表示为
$ Y = \frac{1}{\alpha }(X - m). $
因此,类似于上述摩尔纹生成网络,摩尔纹去除问题也可以分为2个子问题:摩尔纹伪影去除和全局强度恢复. 给定摩尔纹图像$ \alpha Y + m $,摩尔纹去除模块去除摩尔纹图案$ m $,强度恢复模块估计全局强度恢复参数$ \alpha $.
在EEUCN中,摩尔纹生成网络是在已知$ {G_{\text{D}}} $的情况下训练$ {G_{\text{M}}} $,网络中的鉴别器$ {D_{\text{M}}} $用于区分真实的摩尔纹图像与网络生成的摩尔纹图像;反之,摩尔纹去除网络是在已知$ {G_{\text{M}}} $的情况下训练$ {G_{\text{D}}} $,网络中的鉴别器$ {D_{\text{D}}} $用于区分真实的干净图像与网络去除摩尔纹后的图像. 2个网络的整体结构比较类似,主要区别在于循环颜色一致性是在$ Y $和$ {G_{\text{D}}}({G_{\text{M}}}(Y)) $之间计算的,即所谓的内容损失. 具体地,在摩尔纹生成网络中,循环颜色损失$ {\ell _{{\text{cycle}}}} $被定义为重建的摩尔纹图像和原始摩尔纹图像之间的$ {\ell _1} $范数,由式(22)给出:
$ {\ell _{{\text{cycle}}}} = {E_X}[{\left\| {{G_{\text{M}}}({G_{\text{D}}}(X)) - X} \right\|_1}]. $
不同于生成网络,额外的循环损失ASL运算被用于训练摩尔纹去除网络,由式(23)给出:
$ {\ell }_{ASL}={E}_{Y}\left[{\displaystyle \sum _{i=1}^{4}{\Vert {S}_{i}({G}_{\text{D}}({G}_{\text{M}}(Y)))-{S}_{i}(Y)\Vert }_{1}}\right]\text{,} $
其中$ {S_i}( \cdot ) $表示水平、垂直和2个对角滤波器中Sobel滤波中第i个滤波器得到的边缘图. 因此,式(23)中的$ {\ell _{ASL}} $量化了重建图像和原始干净图像的边缘图之间的循环颜色一致性. 由于自然图像包含有意义的边缘信息,ASL运算用于去除已被检测为虚假边缘的摩尔纹伪影. 由于摩尔纹图像的边缘图包含大量噪声,对应于摩尔纹伪影,很难区分真实的和生成的摩尔纹图像的边缘图. 因此,EEUCN只使用 ASL运算来训练摩尔纹去除网络.
摩尔纹去除网络模型的训练数据集质量在很大程度上影响了深度学习方法的处理性能,因此训练数据的收集和对齐工作也至关重要. 现实生活中人们通常会通过屏幕拍照和屏幕录像的方式获取屏摄图像,而将视频文件一般转化为带有摩尔纹图案的图像帧序列,因此本文主要聚焦于屏摄图像中的摩尔纹图案去除研究. 目前摩尔纹图像训练数据集的构建主要从正向和反向2个角度展开. 所谓正向构建是指通过手机设备拍摄电子屏幕得到包含摩尔纹的屏摄图像,然后利用图像处理方法将屏摄图像进行矫正预处理,并将预处理后的摩尔纹图像与原始的干净图像对齐形成图像对,例如TIP2018数据集[40]与MRBI数据集[67]. 反向构建方式是指利用马赛克采样、重上色等操作处理原始的干净图像得到被污染的图像,用于模拟摩尔纹图案生成,从而形成干净图像和摩尔纹图像的图像对,例如在多个国际大型的摩尔纹图像去除竞赛中用到的LCD Moiré数据集[68]及CFA Moiré数据集[69]. 表2为4种数据集间的对比和总结,下面分别对这4种训练数据集的构造过程进行简要介绍.
TIP2018数据集主要包含135 000个图像对,每个图像对包含1幅被摩尔纹污染的图像及其对应的未污染参考图像,其中未受污染的参考图像来自ImageNet ISVRC 2012数据集的100 000幅验证图像和50 000幅测试图像. 在所有图像中,90%用作训练集,10%用于验证和测试. 收集这些数据的流程如图6所示,主要包括2个步骤:图像捕获和图像对齐. 1)图像捕获. 如图6(a)所示,每个参考图像T用黑色边框增强并显示在计算机屏幕的中心位置,以最大程度减小受摩尔纹效应的影响. 为了增加图像对齐过程中可以使用的角点数量,该方法进一步从黑色边框的每条边挤出一个黑色块,并用纯白色填充黑色边框(和块)之外的屏幕其余部分,这使拍摄者能够轻松地检测捕获图像中的黑色边框. 在图像采集过程中,拍摄者使用不同型号的智能手机和显示器屏幕设备,并随机改变手机和计算机屏幕之间的距离和角度,确保在不同的光学传感器上捕捉到的摩尔纹图案S具有多样性. 为了提高图像捕获效率,将参考图像T按照一定的时间间隔在屏幕上连续显示,并使用手机记录连续显示图像的视频,然后将拍摄视频中的帧提取为摩尔纹图像. 2)图像对齐. 实际上,该方法依靠沿着黑色图像边界的角点来完成图像对齐,捕获图像S和参考图像T中的对应点通过单应性关联,并用具有 8个自由度的 3×3 投影矩阵表示. 为了提高配准精度,附加到图像边界的4个黑色块将非共线对应点的数量从4个增加到20个,利用这些角点来计算投影矩阵,并进一步对齐每对图像. 为了检测角点,屏摄图像首先进行二值化处理,利用传统Harris角点检测方法沿黑色图像边界的最外边界搜索角点. 然而,由于摩尔纹伪影的存在,图6(b)所示的20个角点有时无法可靠地被检测到,其中某些边缘像素可能被错误地检测为角点. 经过额外的假角点和重复角点消除处理后,所有的20个角点都可以被成功检测到,如图6(c)所示. 最后,通过计算出的投影矩阵,该方法可以对齐每个图像对,配准结果如图6(c)(f)所示. TIP2018数据集是第一个被大规模用于摩尔纹去除网络的训练数据集,后续很多工作都是基于该数据集进行算法优化和对比评测.
文献[67]指出TIP2018数据集在视频模式下录制往往会使重新捕获的帧变亮,这使得图像对之间的亮度差异很小,这与实际的屏幕拍摄图像效果不符. 此外,为了应用图像对齐算法,TIP2018数据集中的参考图像被要求显示在平板显示器的屏幕中心位置(以能够拍摄全部的黑色边区域),这也限制了摩尔纹图像的种类. 为了克服上述缺陷,文献[67]通过5种不同型号的智能手机拍摄3种不同类型的显示器屏幕如笔记本电脑、台式电脑的LED显示屏以及电视的LCD屏来构建MRBI数据集,其训练集包含352幅重新捕获的带有摩尔纹伪影的屏幕图像和相应的原始图像(通过屏幕截图获得). 测试集包含由 24 种屏幕和相机组合捕获的 340 幅图像,它们代表了各种各样的摩尔纹图像. 具体地,在创建训练图像对时,首先对相同内容的屏幕分别通过屏幕截屏和屏幕拍照得到干净图像$ {I_{\text{o}}} $和包含摩尔纹的图像$ {I_{\text{m}}} $;接着对$ {I_{\text{m}}} $缩放使$I_{\mathrm{o}} $和$I_{\mathrm{m}} $这2幅图像大小相似;然后利用基于尺度不变的特征变换(scale-invariant feature transform, SIFT)[70]找到2幅图像中可以匹配的一系列点构成点对的集合$ {{\varOmega }} $;然后,随机从$ {{\varOmega }} $中选出4个点对,用RANSAC算法[71]估计$ {I_{\text{o}}} $和$ {I_{\text{m}}} $之间的变换矩阵$ {\boldsymbol{H}} $,$ {{\varOmega }} $中的其他点对用于验证$ {\boldsymbol{H}} $的准确性. 验证方式是点$ x_i^{\text{m}} $与$ {\boldsymbol{H}}x_i^{\text{0}} $的距离是否小于预先设置的阈值$ \varepsilon $,若是,则点对$ x_i^{\text{m}} $与$ x_i^{\text{0}} $被称之为内点. 重复上述过程直到所求的变换矩阵$ {\boldsymbol{H}} $满足使$ {{\varOmega }} $中内点的数目最多. 最后对$ {I_{\text{o}}} $做矩阵为$ {\boldsymbol{H}} $的变换得到一个训练图像对. 通过上述方法构造的MRBI数据集为图像去摩尔纹和亮度改善提供了基准,并将会激发更多关于这一主题的工作.
LCD Moiré数据集[68]包含10 200个合成生成的图像对(由摩尔纹退化的图像和干净的原始图像组成,10 000对用于网络训练、100对用于验证及100对用于测试),是专门为举办首次图像去摩尔纹挑战赛而创建的. 该挑战赛是与ICCV 2019联合举办的图像处理进展(Advances in Image Manipulation,AIM)研讨会的一部分. LCD Moiré数据集中的干净图像是从ICCV(IEEE International Conference on Computer Vision),ECCV(European Conference on Computer Vision),CVPR(IEEE Conference on Computer Vision and Pattern Recognition)等计算机视觉会议论文中收集,由文本、图形或等比例组合的文本/图形组成. 摩尔纹图像的生成相当于模拟使用智能手机在显示干净图像的LCD屏幕上拍照的过程. 具体的思路为:经过RGB图像的马赛克重采样处理,以模拟LCD屏幕显示图像;对图像应用随机投影变换以模拟显示器和相机的不同相对位置和方向,并应用3×3高斯滤波器;使用Bayer CFA对图像重新采样以模拟原始数据,并添加高斯噪声以模拟传感器噪声;应用简单的ISP处理,包括去马赛克(双线性插值)和去噪;使用JPEG压缩来压缩图像以模拟压缩噪声,将干净图像与摩尔纹图像对齐,并裁剪出图像对. 为了确保训练、验证和测试图像集中的内容和摩尔纹图案分布相同,该方法附加了图像内容平衡和摩尔纹分量平衡等后处理操作.
2020年,CVPR的图像恢复增强竞赛(NTIRE)提供了一个新的摩尔纹图像训练数据集CFA Moiré Dataset[69],该数据集由10 000个训练图像对、500个验证图像对和500个测试图像对组成. 其中干净的图像从现有的图像数据库中采集或者裁剪后得到,图像质量较高,内容涵盖了服装、建筑等高频重复图案. 相应的摩尔纹图像通过马赛克重采样和去马赛克过程生成,其中摩尔纹伪影图案出现在图像高频区域. 数据集构建分为2步:首先,通过傅里叶频域中摩尔纹伪影的量化自动选择符合条件的图像对,即测量清洁图像和去马赛克图像之间的频率变化;然后,通过人工选择进一步优化选择的图像对,将已经被污染的干净图像从数据集中删除.
随着不同拍摄设备和显示设备的广泛普及,摩尔纹图像的呈现也趋于多样化. 由于摩尔纹图案不规则,且不均匀地分布在图像中的不同空域及不同频域段中,其分布特征无法通过先验知识准确获取. 另外,当摩尔纹分量与图像中的纹理分量出现混叠时,基于先验知识的图像处理方法无法将其准确区分,因此基于特定先验知识的传统图像处理方法在摩尔纹去除方面愈加受限. 而在训练数据集构建逐渐完善的情况下,深度学习方法的处理性能更加显著,并逐渐成为主流方法. 目前的神经网络模型主要利用CNN卷积神经网络模型和GAN对抗网络,通过提取空域或频域中的特征信息达到摩尔纹去除的目的. 本节主要将基于深度学习的摩尔纹图案去除方法进行简单回顾总结,并分别对各种方法的优缺点进行定性的对比分析. 另外基于相同的公开数据集,我们选取了部分代表性工作进行了算法实现,并给出了定量的性能对比结果.
首先,在基于CNN空域算法中,DMCNN[40]模型和DCNN[41]较早地将深度学习方法应用于摩尔纹图像去除,并实现了不同分辨率分支下的特征提取. 由于摩尔纹不规则地分布在图像的多个频段中,因此,提取图像的多尺度特征信息成了学者们的共识. 但是此类方法在每一个尺度上提取的特征过于简单,且忽略了不同尺度特征之间的联系,因此具有一定的局限性. 例如:当图像中出现大面积彩色条纹分布的摩尔纹时,摩尔纹去除的效果不理想. MSFE[42]模型是基于DMCNN提出的,其明显的缺陷是仅仅将低分辨率的特征嵌入到了高分辨率特征中,而高分辨率的特征没有融合进低分辨率的特征中,因此对比DMCNN模型优势不明显. 相对于DMCNN而言,MopNet[43]考虑了更多的摩尔纹图案的不同属性特征,但是没有综合分析这些特征在多个不同尺度间的联系,而且针对边界的提取无法获得复杂的纹理特征. 因此,MopNet无法准确区分出污染图像中的纹理区域和摩尔纹区域. HRDN[44]在上采样和下采样的过程中针对不同的尺度做了较为充分的信息交换,但每个尺度中提取的特征信息不够充分. MDDM[45]模型的分层数目过多,各个尺度之间仍然缺乏信息的交互,同样存在低分辨率分支中没有融合进高分辨率信息的现象. 并且高分辨率的分层卷积数目少,低分辨率的分层卷积数目多,理论上应该反过来,因为高分辨率的分支拥有更多的特征. 相较于MDDM,改进后的模型MDDM+[46]不仅视觉效果得到明显的改善,而且存储的参数数量由8.01 MB变为3.57 MB,实用性进一步提高. 但相较于HRDN模型,MDDM+仍存在各分支间缺少信息交互的特点.
在基于CNN频域算法中, MBCNN[47]在提取摩尔纹分量DCT频域特征时,每个摩尔纹去除模块可以根据多尺度频域特征采用不同的权值. 相对于MBCNN,MBCNN+[60]在网络结构设计和损失函数中改进了不同尺度特征提取和融合的方式,因此取得更好的输出图像视觉效果. WDNet[64]则是基于小波域实现,与傅里叶变换和离散余弦变换相比,小波变换既考虑了空间域信息,又考虑了频域信息. 在小波合成中,不同的小波带代表如此广泛的频率范围,这是几个卷积层即使是大核也无法实现的.
基于GAN的摩尔纹图案去除方法是最近2年提出的,较好地解决了CNN方法依赖大量对齐的训练对的问题. 该类方法的处理性能跟CNN方法相比还有一定的差距,但为摩尔纹图案去除方法研究提供了一种新的思路.
综上所述,不同的网络模型架构具有各自的优势和劣势,具体的对比分析如表3所示.
在本节中,我们基于相同的公开的数据集,分别选择不同类别中有代表性的方法进行了算法实现和性能对比实验. CNN空域算法模型选择:首次使用的DMCNN[40]训练模型;融合多尺度特征聚合、信道方向目标边缘预测器和模式分类器的MopNet[43]训练模型;不同尺度间特征能够较充分融合的HRDN[44]训练模型. 在CNN频域算法模型方面,我们选取了基于DCT频域特征信息提取的MBCNN[47]训练模型. 相应地,在GAN网络模型中,我们选取了目前性能最好的EEUCN[50]模型. 测试环境为torch1.8.0,GPU3090,数据集选用TIP2018数据集. 在摩尔纹图案去除算法性能的评估方面,我们采取了类似ICCV 2019 AIM[69]和2020 CVPR NTIRE[70]竞赛的评判规则,分别计算了在TIP2018数据集上取得平均的PSNR值和SSIM值.
首先,在模型预处理阶段,我们在Sun等人[40]预处理方法的基础上,对TIP2018数据集中的摩尔纹图像进行了重采样和归一化处理. 为了保证每个模型训练数据的一致性和公平性,我们加载了MopNet[43]的预训练模型对上述数据再次进行预处理操作,预处理后的图像样本尺寸统一为256×256像素. 据此,我们分别针对不同的网络模型采取不同的训练方式.
1) DMCNN方法. 使用Adam优化器,学习率设置为0.0001,当损失不下降时缩减学习率为原来的90%,共计训练50个epoch.
2) MopNet方法. 训练模式分为2个阶段: 第1阶段,训练边缘提取模型和分类模型直到模型收敛,2个模型分别训练50个epoch和20个epoch;第2阶段,使用Adam优化器,学习率最初设置为0.0002,并随着训练线性衰减,共计训练150个epoch.
3) HRDN方法. 使用Adam优化器,学习率设置为0.0001,当损失不下降或每训练10个epoch时下降为原来的50%,共计训练50个epoch.
4) MBCNN方法. 使用Adam优化器,学习率设置为0.0001,共计训练100个epoch.
5) EEUCN方法. 使用AdamW优化器,学习率设置为0.0001,在训练100个epoch后,学习率缩减为原来的10%,共计训练150个epoch. 由于EEUCN是使用unpair数据集的方法,我们基于文献[50]提供的LCD Moire上非成对的数据集进行训练,并使用TIP2018数据集进行测试.
不同的训练模型在TIP2018数据集上取得平均的PSNR值和SSIM值如表4所示,其中部分图像在不同模型上得到的去除摩尔纹图像效果如图7所示.
如表4和图7所示,DMCNN[40]是利用多尺度卷积神经网络去除摩尔纹图案的初步尝试,在摩尔纹分量的去除和图像色调恢复方面,相比其他模型而言,该模型输出结果不太理想.EEUCN[50]是利用GAN网络去除摩尔纹图案的尝试,其思想是先利用摩尔纹生成网络在干净的图像中添加摩尔纹图案,然后利用监督学习方法去除摩尔纹图案. 由于模拟生成的摩尔纹图案与真实的摩尔纹图案之间存在一定的差异,当训练数据集和测试数据集不同时,该模型输出的去摩尔纹图像存在较大的颜色失真,例如图7(g)所示的第8幅图像中的多个颜色分量被误删除. 诸如MopNet[43],HRDN[44],MBCNN[47]等有监督的多尺度卷积神经网络则可以得到较理想的图像质量. 但由于MopNet仅通过边界提取区分摩尔纹分量与图像分量,导致如图7(d)中第4幅图像的前景区域、第5幅图像的人脸区域、第6幅图像的翅膀区域、第9幅图像的瓢虫区域等局部摩尔纹图案去除不干净,另外第8图像中的复杂区域则产生严重的色调失真. HRDN在特征提取和融合阶段具有相对充分的信息交换过程,在TIP2018数据集中取得较高的PSNR值和SSIM值. 但由于该模型中各尺度特征提取不够充分,当处理包含特殊的复杂摩尔纹图案时,HRDN方法处理的图像视觉效果不很理想,如图7(e)中的第2幅、第6幅中出现了形状复杂的彩色摩尔纹图案,HRDN未能准确提取出摩尔纹分量. MBCNN通过提取摩尔纹图像的频域特征,并进行频域中摩尔纹分量去除,同时融入了色调恢复过程,则在有监督学习的卷积神经网络模型中取得了较为理想的视觉效果. 其中SSIM值明显高于其他模型,则说明该模型输出去除摩尔纹图像在亮度、对比度、结构上与原始干净图像保持较好的一致性,如图7(f)所示,MBCNN能够较好地处理各种复杂特殊的摩尔纹图案.
图像去摩尔纹是一项重要的图像修复任务,其目的是从污染图像中去除摩尔纹干扰图案并恢复底层干净的图像. 传统的基于先验知识的方法具有算法效率高、不需要大规模训练等优点,能够较好地处理频域或空域分布特征相对简单的摩尔纹图案. 比如当利用智能手机或者数码相机拍摄布料、织物、瓷砖等高频重复纹理静态图案时,在频域中摩尔纹的能量是集中的,与图像的纹理能量有一定的差距. 因此,基于图像分解的方法则可以更好地分离2种不同能量,进而有效去除摩尔纹图案. 对于扫描的半色调图像而言,摩尔纹图案可能会呈现周期性的网纹干扰性质,利用传统图像滤波的方法性能可能会更好;基于图像差值的方法在去除摩尔纹图案的同时,可以更好地恢复图像亮度信息和色彩空间信息.
然而,与图像去噪任务中均匀分布的噪声和超分辨率任务中缺失的高频细节不同,屏幕拍摄的摩尔纹图案具有广泛分布的频谱和复杂纹理的动态特性,这使得图像去摩尔纹任务面临着巨大挑战. 传统的基于信号处理的摩尔纹去除方法需要基于特定的先验知识和模型,因此图像修复效果存在较大的提升空间. 基于深度学习的摩尔纹去除方法则通过在不同尺度上探究摩尔纹图案和背景图像之间的内在相关性,并考虑不同尺度之间的特征融合,在图像质量改善方面取得较好的性能. 但是基于CNN的网络模型需要大规模的摩尔纹污染图像和原始干净图像的训练集对的支撑,数据集构造的质量也会严重影响算法性能,而相关数据集对的构造难度较大,在很多情况下无法完成. 综上,针对目前摩尔纹图案去除方法存在的问题,为了进一步提高网络模型的实用性,以下3个技术方向值得进一步关注.
1) 摩尔纹成因描述模型构建
在早期阶段,人们对于摩尔纹的研究大多基于数学模型,从摩尔纹的成因出发,研究摩尔纹的频域和空域特征. 不同的成像模式也会造成摩尔纹特性表现的差异,通过对摩尔纹成因的深入研究,我们可以构建更为准确的摩尔纹消除模型. 一方面,摩尔纹成因分析有利于特定场景下的摩尔纹图案去除. 刘芳蕾[21]通过对纹理图像中摩尔纹分量的频域稀疏特性和纹理分量的空余低秩特性的分析,构建图像分解模型实现了纹理层与摩尔纹的分离. 张雪[22]针对屏摄图像中摩尔纹能量分布非常分散、高频信息难以区分等特点,并结合屏摄摩尔纹和图像背景的结构性差异,提出了基于GMM的多相图像分解方法,在保留背景图像结构的同时扰乱摩尔纹结构,从而避免块效应并降低算法的时间复杂度. 另一方面,精确的摩尔纹成因描述模型也可以辅助深度学习模型的训练,促进网络结构优化,提升摩尔纹图案消除算法性能. 鉴于摩尔纹图案的复杂特性和成因差异,现有描述摩尔纹成因的数学模型缺乏通用性,需要未来进一步的深入研究.
2) 真实场景下的通用摩尔纹去除网络研究
目前基于深度学习的摩尔纹图案去除网络大都是采用实验室环境下采集和模拟生成的数据集进行训练和评估的,在某些情况不能完全适用于现实场景中具有复杂特征的污染图像,模型泛化性能较低. 由表2可知,规模最大的TIP2018数据集最接近于真实场景下的摩尔纹图案效果,但其图像采集过程也存在图像分辨率低、摩尔纹场景简单、拍摄角度单一等问题. 而实际需要处理的摩尔纹图案往往具有分辨率高,且摩尔纹颜色、形状、边界、密度等复杂多变的特点. 为此,我们需要从以下方面展开深入研究,以探索真实场景下的摩尔纹图案去除方案:①利用人工智能和计算机视觉技术,在真实场景下构建更大规模的摩尔纹图案数据集,包含不同分辨率、不同场景、不同摩尔纹强度等多种特征;②充分探索不同分辨率特征映射之间的关系,将不同尺度下的摩尔纹颜色、形状、边界、密度等多种特征信息进行融合和交换,然后通过学习和解耦得到更加通用的摩尔纹图案相关语义特征,而忽略无关特征,从而构建高效和高鲁棒的通用摩尔纹去除网络模型结构;③基于对抗神经网络GAN的训练模型,在网络生成器和鉴别器中引入基于多尺度特征融合的CNN卷积神经网络,将会成为摩尔纹去除技术的一个重要研究方向.
3) 基于移动终端的泛化应用拓展
随着移动互联网技术的飞速发展和智能移动终端设备的全面普及,利用智能移动终端进行信息捕获、处理、存储和共享等成为主流应用方式,比如图像美化[72]、OCR识别[73]和水印提取[74]等. 由于屏摄图像中通常会受到摩尔纹干扰图案的严重干扰,基于移动终端的摩尔纹去除应用存在诸多不足之处:①硬件设备的多样性和不同设备之间的差异性导致了摩尔纹图案的频域特征分布更加不规则,深度学习网络模型的泛化性能普遍较低;②现有的基于深度学习的摩尔纹去除网络模型大多都是部署在服务器端,其网络结构复杂、训练数据庞大、计算资源配置高,较难直接部署于计算能力有限的移动终端设备;③深度学习算法与移动终端应用结合的研究有待深入,鲜有直接应用于移动终端的轻量级深度学习模型被提出,基于移动终端的摩尔纹去除方法实时性较差. 为此,需要在2个方面进行基于移动终端的摩尔纹去除网络模型的优化和构建:①利用低秩近似分解、网络剪枝和网络量化对已有的深度学习网络进行模型压缩和加速优化;直接设计轻量化的网络结构,并将训练完成的深层网络模型完全运行于移动终端设备. ②结合边缘计算技术,通过神经网络前向推理框架[75]将优化后的深度学习模型部署至移动端的应用系统.
总之,基于深度学习的摩尔纹图案去除模型的部署轻量化、计算低配化、性能普适化以及操作实时化将成为移动终端泛化应用的重要研究目标.
作者贡献声明:亓文法提出了文章框架结构和研究思路,并撰写论文;刘宇鑫负责完成实验并撰写论文;郭宗明提出指导意见并修改论文.

参考文献 引证文献
排序方式:
1
Liu Shuai, Li Chenghua, Nan Nan, et al. MMDM: Multi-frame and multi-scale for image deMoiréing[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway, NJ: IEEE, 2020: 1751−1759
2
Niu Wenjia, Zhang Kaihao, Luo Wenhan, et al. Blind motion deblurring super-resolution: When dynamic spatio-temporal learning meets static image understanding[J]. IEEE Transactions on Image Processing, 2021, 30(8): 7101−7111
3
Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C] //Proc of the 18th 2015 Medical Image Computing and Computer-Assisted Intervention. Berlin: Springer, 2015: 234–241
4
Wang Fuping, Hu Yang, Liu Weihua, et al. Face impainting algorithm combining face sketch and gate convolution[C]//Proc of 4th Int Conf on Natural Language Processing(ICNLP). Piscataway, NJ: IEEE, 2022: 81−86
5
蒋斌,刘虹雨,杨超,等. 一种基于局部属性生成对抗网络的人脸修复算法[J]. 计算机研究与发展,2019,56(11):2485−2493
Jiang Bin, Liu Hongyu, Yang Chao, et al. A face inpainting algorithm with local attribute generative adversarial networks[J]. Journal of Computer Research and Development, 2019, 56(11): 2485−2493(in Chinese)
6
Gao Tianyu, Guo Yanqing, Zheng Xin, et al. Moiré pattern removal with multi-scale feature enhancing network[C] //Proc of IEEE Int Conf on Multimedia and Expo Workshops (ICMEW). Piscataway, NJ: IEEE, 2019: 240−245
7
Elad M, Aharon M. Image denoising via sparse and redundant representations over learned dictionaries[J]. IEEE Transactions on Image Processing, 2006, 15(12): 3736−3745
8
Chen Xiaogang, Kang Singbing, Yang Jie, et al. Fast patch-based denoising using approximated patch geodesic paths[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2013: 1211–1218
9
Li Xu, Qiong Yan, Yang Xia, et al. Structure extraction from texture via relative total variation[J]. ACM Transactions on Graphics, 2012, 31(6): 139−148
10
Shou Yuwen, Lin Chinteng. Image descreening by GA-CNN-based texture classification[J]. IEEE Transactions on Circuits and Systems I: Regular Papers, 2004, 51(11): 2287−2299
11
Nah S, Kim T H, Lee K M, Deep multi-scale convolutional neural network for dynamic scene deblurring[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017: 3883–3891
12
Nishioka K, Hasegawa N, Ono K, et al. Endoscope system provided with low-pass filter for Moiré removal: US, US08/917429[P]. 2000-02-15
13
Schoberl M, Schnurrer W, Oberdorster A, et al. Dimensioning of optical birefringent anti-alias filters for digital cameras[C] //Proc of IEEE Int Conf on Image Process (ICIP). Piscataway, NJ: IEEE, 2010: 4305–4308
14
Menon D, Calvagno G. Color image demosaicking: An overview[J]. Signal Processing: Image Communication, 2011, 26(8/9): 518−533
15
Pekkucuksen I, Altunbasak Y. Multiscale gradients-based color filter array interpolation[J]. IEEE Transactions on Image Processing, 2013, 22(1): 157−165
16
邱菊. 基于插值算法的去彩色摩尔纹效应的研究[D]. 南京:南京理工大学,2009
Qiu Ju. Research on color Moiré removal based on interpolation algorithm [D]. Nanjing: Nanjing University of Science and Technology, 2009 (in Chinese)
17
邱香香. 基于CCD静态图像的摩尔纹去除算法研究[D]. 南京:南京理工大学,2012
Qiu Xiangxiang. Research on Moiré removal algorithm based on CCD static image[D]. Nanjing: Nanjing University of Science and Technology, 2012 (in Chinese)
18
Hibbard R H. Apparatus and method for adaptively interpolating a full color image utilizing luminance gradients: US, US08/085520[P]. 1995-01-17
19
Hamilton J F. Adaptive color plane interpolation in single sensor color electronic camera: US, US08/407423[P]. 1997-05-13
20
Kimmel R. Demosaicing: Image reconstruction from color CCD samples[J]. IEEE Transactions on Image Processing, 1999, 9(8): 1221−1228
21
刘芳蕾. 基于图像分解的纹理图像摩尔纹消除方法[D]. 天津:天津大学,2017
Liu Fanglei. Moiré pattern removal from Textured images via image decomposition[D]. Tianjin : Tianjin University, 2017 (in Chinese)
22
张雪. 基于图像分解的屏摄图像摩尔纹消除方法[D]. 天津:天津大学,2018
Zhang Xue. Moiré pattern removal from scree-shot images via image decomposition[D]. Tianjin : Tianjin University, 2018 (in Chinese)
23
Liu Fanglei, Yang Jingyu, Yue Huanjing. Moiré pattern removal from texture images via low-rank and sparse matrix decomposition[C/OL] //Proc of IEEE Int Conf on Visual Communications and Image Processing (VCIP). Piscataway, NJ: IEEE, 2015[2016-08-25].https:// ieeexplore. ieee. org/stamp/stamp. jsp?tp=&arnumber=7457907
24
Yang Jingyu, Zhang Xue, Cai Changrui, et al. DeMoiréing for screen-shot images with multi-channel layer decomposition[C/OL] //Proc of IEEE Visual Communications and Image Processing (VCIP). Piscataway, NJ: IEEE, 2017[2018-03-01].https://ieeexplore.ieee.org/stamp/stamp.jsp?tp =& arnumber=8305057
25
Fang Faming, Wang Tingting, Wu Shuyan, et al. Removing Moiré patterns from single images[J]. Information Sciences, 2020, 514(8): 56−71
26
Sasada R, Yamada M, Hara S, et al. Stationary grid pattern removal using 2D technique for Moiré-free radiographic image display[C] //Proc of Conf on Medical Imaging: Visualization, Image-Guided Procedures, and Display. Bellingham, WA: SPIE, 2003: 688–697
27
Sidorov D N, Kokaram A C, Suppression of Moiré patterns via spectral analysis[C] //Proc of Visual Communications and Image Processing. Bellingham, WA: SPIE, 2002: 895–906
28
Wei Zhouping, Wang Jian, Nichol H, et al. A median-Gaussian filtering framework for Moiré pattern noise removal from X-ray microscopy image[J]. Micron, 2012, 43(2/3): 170−176
29
Hazavei S M, Shahdoosti H R. A new method for removing the Moiré pattern from images[J]. arXiv preprint, arXiv: 1701. 09037, 2017
30
Garcia D C, Queiroz R L. Face-spoofing 2D-detection based on Moiré-pattern analysis[J]. IEEE Transactions on Information Forensics and Security, 2015, 10(4): 778−786
31
Sidorov D N, Kokaram A C. Removing Moiré from degraded video archives[C/OL] //Proc of the 11th European Signal Processing Conf. Piscataway, NJ: IEEE , 2002[2015-03-30].https://ieeexplore.ieee.org/ stamp/stamp.jsp?tp=&arnumber=7071920
32
吕旭,胡柏青,李开龙,等. 非线性目标跟踪滤波算法的比较研究[J]. 火力与指挥控制,2021,46(4):24−30
Lü Xu, Hu Baiqing, Li Kailong, et al. Comparison of nonlinear target tracking filtering algorithm[J]. Fire Control & Command Control, 2021, 46(4): 24−30(in Chinese)
33
Sun Bin, Li Shutao, Sun Jun. Scanned image descreening with image redundancy and adaptive filtering[J]. IEEE Transactions on Image Processing, 2014, 23(8): 3698−3710
34
Dong Chao, Loy Chenchang, He Kaiming, et al. Learning a deep convolutional network for image super-resolution[C] //Proc of the 13th European Conf on Computer Vision. Berlin: Springer, 2014: 184–199
35
Kim J, Lee J K, Lee K M, Convolutional networks for biomedical using very deep convolutional networks[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2016: 1646–1654
36
Gharbi M, Chaurasia G, Paris S, et al, Deep joint demosaicking and denoising[J], ACM Transaction of Graphics, 2016, 35(6): 1−12
37
Zhang Kai, Zuo Wangmeng, Chen Yujin, et al. Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3142−3155
38
Zhang Kai, Zuo Wangmeng, Chen Yujin, et al. Learning deep CNN denoiser prior for image restoration[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017: 2808–2817
39
Abraham E. Moiré pattern detection using wavelet decomposition and convolutional neural network[C] //Proc of IEEE Symp Series on Computational Intelligence (SSCI). Piscataway, NJ: IEEE, 2018: 1275−1279
40
Sun Yujing, Yu Yizhou, Wang Wenping. Moiré photo restoration using multiresolution convolutional neural networks[J]. IEEE Transactions on Image Processing, 2018, 27(8): 4160−4172
41
Liu Bolin, Shu Xiao, Wu Xiaolin. DeMoiréing of camera-captured screen images using deep convolutional neural network[J]. arXiv preprint, arXiv: 1804. 03809, 2018
42
Gao Tianyu, Guo Yanqing, Zheng Xin, et al. Moiré pattern removal with multi-scale feature enhancing network[C] //Proc of IEEE Int Conf on Multimedia and Expo Workshops. Piscataway, NJ: IEEE, 2019: 240−245
43
He Bin, Wang Ce, Shi Boxin, et al. Mop Moiré patterns using mopnet[C] //Proc of IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2019: 2424−2432
44
Yang Shanghui, Lei Yajing, Xiong Shuangyu, et al. High resolution deMoiré network[C] //Proc of IEEE Int Conf on Image Processing (ICIP). Piscataway, NJ: IEEE, 2020: 888-892
45
Cheng Xi, Fu Zhenyong, Yang Jian. Multi-scale dynamic feature encoding network for image deMoiréing[C] //Proc of IEEE/CVF Int Conf on Computer Vision Workshop (ICCVW). Piscataway, NJ: IEEE, 2019: 3486-3493
46
Cheng Xi, Fu Zhenyong, Yang Jian. Improved multi-scale dynamic feature encoding network for image deMoiréing[J]. Pattern Recognition, 2021, 116(1): 1−12
47
Zheng Bolun, Yuan Shanxi, Slabaugh G, et al. Image deMoiréing with learnable bandpass filters[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2020: 3636–3645
48
Park H, Vien A G, Koh Y J, et al. Unpaired image deMoiréing based on cyclic Moiré learning[C] //Proc of Asia–Pacific Signal Information Processing Association Annual Summit and Conf. Piscataway, NJ: IEEE, 2021: 146–150
49
Yue Huanjin, Cheng Yijia, Liu Fanglong, et al. Unsupervised Moiré pattern removal for recaptured screen images[J]. Neurocomputing, 2021, 456: 352−363
50
Park H, Vien A G, Kim H, et al. Unpaired screen-shot image deMoiréing with cyclic Moiré learning[J]. IEEE Access,2022,10:16254-16268
51
Yang Jingyu, Liu Fanglei, Yue Huanjing, et al. Textured image deMoiréing via signal decomposition and guided filtering[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3528−3541
52
Ok J, Youn S, Seo G, et al. Paper check image quality enhancement with Moiré reduction[J]. Multimedia Tools and Applications, 2017, 76(20): 21423−21450
53
Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[J]. arXiv preprint, arXiv: 1505. 04597v1, 2015
54
Gao Huang, Zhuang Liu, Laurens V D M, et al. Densely connected convolutional networks[J]. arXiv preprint, arXiv: 1608. 06993v5, 2018
55
Hu Jie, Shen Li, Sun Gang. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011−2023
56
Huang Xun, Belongie S. Arbitrary style transfer in real-time with adaptive instance normalization[C] //Proc of IEEE Int Conf on Computer Vision (ICCV). Piscataway, NJ: IEEE, 2017: 1501–1510
57
Vien A G, Park H, Lee C. Dual-domain deep convolutional neural networks for image deMoiréing[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway, NJ: IEEE, 2020: 1934−1942
58
Vien A G, Park H, Lee C. Moiré artifacts removal in screen-shot images via multiple domain learning[C] //Proc of Asia-Pacific Signal and Information Processing Association Annual Summit and Conf (APSIPA ASC). Piscataway, NJ: IEEE, 2020: 1268−1273
59
Wang Hailing, Tian Qiaoyu, Li Liang, et al. Image deMoiréing with a dual domain distilling network[C/OL] //Proc of IEEE Int Conf on Multimedia and Expo (ICME). Piscataway, NJ: IEEE, 2021[2021-06-09]. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9428091
60
Zheng Bolun, Yuan Shanxi, Yan Chengganng, et al. Learning frequency domain priors for image deMoiréing[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(11): 7705−7717
61
Liu Lin, Liu Jianzhuang, Yuan Shanxin, et al. Wavelet-based dual-branch network for image deMoiréing[J]. arXiv preprint, arXiv: 2007. 07173v2, 2020
62
Gao Huang, Zhuang Liu, Laurens M, et al. Densely connected convolutional networks[C] //Proc of IEEE Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017, 2261−2269
63
Yu F, Koltun V. Multi-scale context aggregation by dilated convolutions[J]. arXiv preprint, arXiv: 1511. 07122, 2016
64
Zheng Yuanjie, Yu Jingyi, Kang Singbing, et al. Single-image vignetting correction using radial gradient symmetry[C/OL] //Proc of IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2008[2008-08-05].https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=& arnumber =4587413
65
Dai Tao, Cai Jianrui, Zhang Yongbing, et al. Second-order attention network for single image super-resolution[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2019: 11065–11074
66
Mao Xudong, Li Qing, Xie Haoran, et al. Least squares generative adversarial networks[C] //Proc of IEEE Int Conf on Computer Vision (ICCV). Piscataway, NJ: IEEE, 2017: 2794–2802
67
Yue Huanjing, Mao Yan, Liang Lipu, et al. Recaptured screen image deMoiréing[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2021, 31(1): 49−60
68
Yuan Shanxin, Timofte R, Gregory S, et al. Aim 2019 challenge on image demoreing: Dataset and study[C] //Proc of IEEE Int Conf on Computer Vision (ICCV) Workshops. Piscataway, NJ: IEEE, 2019: 3526–3533
69
Yuan Shanxin, Timofte R, Leonardis A, et al. NTIRE 2020 challenge on image demoireing: Methods and results[C] //Proc of IEEE/CVF Conf on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway, NJ: IEEE, 2020: 1882-1893
70
Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91–110
71
Chin T J, Yu J, Suter D. Accelerated hypothesis generation for multi-structure robust fitting[C] //Proc of the 11th European Conf on Computer Vision. Berlin: Springer, 2010: 533–546
72
高天宇. 面向移动终端的摩尔纹去除系统的设计与实现[D]. 大连:大连理工大学,2021
Gao Tianyu. Design and implementation of Moiré removal system for mobile terminals[D]. Dalian: Dalian University of Technology, 2021 (in Chinese)
73
Kaur R, Sharma D V. Pre-processing for improving the recognition of mobile based Gurmukhi text recognition system[C/OL] //Proc of the 12th Int Conf on Computing Communication and Networking Technologies (ICCCNT). Piscataway, NJ: IEEE, 2021[2021-11-03]. http:// ieeexplore.ieee.org/stamp/ stamp.jsp?tp=&arnumber=9579867
74
Fang Han, Zhang Weiming, Zhou Huang, et al. Screen-shooting resilient watermarking[J]. IEEE Transactions on Information Forensics and Security, 2019, 14(6): 1403−1418
75
Howard A, Sandler M, Chen Bo, et al. Searching for mobilenetv3 [C] //Proc of IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2019: 1314−1324
2024年第61卷第3期
69
31
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202220797
  • 接收时间:2022-09-06
  • 首发时间:2026-11-17
  • 出版时间:2024-03-02
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2022-09-06
  • 修回日期:2023-05-15
基金
作者信息
    1北京大学王选计算机研究所 北京 100080

通讯作者:

郭宗明()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202220797
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏